完成流程的第4轮测试

handoff/2026-06-24-character-consistency-r5-brief-ownership-handoff.md

@@ -0,0 +1,244 @@
+# R5 Brief-Ownership Formal-Run Handoff
+
+Date: 2026-06-24
+
+Project:
+
+```text
+projects/2026-06-23-在路上
+```
+
+## Status
+
+R4 visual production is complete, but the R4 mechanism test should be treated as failed for the next planning step.
+
+R4 accepted images:
+
+- `slides/s02/s02-v2-image.png`
+- `slides/s03/s03-v2-image.png`
+- `slides/s04/s04-v2-image.png`
+
+User accepted all three images. Image quality is no longer the main concern.
+
+R4 mechanism result:
+
+- Clean child session: `019ef8cf-4ab1-7b41-877b-979dc25b0a76`
+- `Agent S-02`: `019ef8d1-188a-70c3-8a7c-4e915144a276`
+- `Agent S-03`: `019ef8e2-5229-72a3-93c9-16e46a87d3bf`
+- `Agent S-04`: `019ef8ec-8032-7de0-a8b8-0d40d456fbaa`
+- `S-02` took about `13m55s`; user observed child-session context at about `107K` tokens after this first package.
+- User observed total child-session context around `132K` tokens after the second package.
+
+R4 failed the intended mechanism check because brief ownership was wrong:
+
+- `s02-v2-brief.md`, `s03-v2-brief.md`, and `s04-v2-brief.md` were generated by the slide Agents.
+- They were not generated and confirmed by the visible child session before launching each slide Agent.
+- Evidence: each slide Agent return lists the brief and packet under worker-written files.
+- `S-04` also has a sequencing warning: the worker launched before visible-child brief/packet prewrite confirmation.
+
+No overwrite, duplicate worker, extra slide, or review-file defect was observed in R4.
+
+## Read First In R5
+
+Read these before starting work:
+
+```text
+AGENTS.md
+docs/workflows/slide-task-orchestration.md
+docs/workflows/slide-task-templates.md
+projects/2026-06-23-在路上/project.md
+projects/2026-06-23-在路上/execution-plan.md
+projects/2026-06-23-在路上/slides/slides.md
+projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/task-log.md
+```
+
+Use this handoff as the R5 task entrypoint.
+
+## R5 Goal
+
+R5 should answer two questions:
+
+1. Can the visible child session generate slide briefs and packets before launching each one-shot slide Agent?
+2. How many real slide shots can one clean visible child session carry before context pressure makes the workflow unreliable?
+
+R5 is not primarily an image-quality test. Images still need to be generated and user-accepted, but the key test is whether the workflow boundary is correct.
+
+## R5 Hard Rule: Brief Ownership
+
+The visible child session must own brief generation.
+
+For each slide/version:
+
+1. The visible child session scans the slide folder and resolves the next unused version.
+2. The visible child session writes `sNN-vN-brief.md`.
+3. The visible child session writes `sNN-vN-packet.md`.
+4. The visible child session confirms both files exist and target paths do not collide.
+5. Only then may it launch the one-shot slide Agent.
+6. The slide Agent reads the brief and packet, then writes prompt/image/return.
+7. The slide Agent must not create or rewrite the brief or packet.
+
+If the worker writes, rewrites, or claims to have written `sNN-vN-brief.md` or `sNN-vN-packet.md`, R5 should return `warning` or `failed` even if the image is acceptable.
+
+## R5 Completion Gate Additions
+
+In addition to the existing completion gate, the child session must verify:
+
+- brief and packet were created by the visible child session before worker launch;
+- worker return does not list brief or packet under worker-written files;
+- worker packet/return timestamps are after visible-child brief/packet timestamps;
+- exactly one owning worker exists per slide/version;
+- no review files exist when review is disabled;
+- generated image path and hash are recorded;
+- child context usage is reported if visible in the runtime, or left for user observation if not available to the child.
+
+If any item fails, report `warning`, `failed`, or `blocked`; do not compress it into a pass.
+
+## Recommended R5 Scope
+
+Use the next natural unfinished sequence after accepted `S-05` and `S-06`:
+
+```text
+S-07
+S-08
+S-09
+```
+
+Expected new versions if no new files appear before R5:
+
+```text
+s07-v2
+s08-v2
+s09-v2
+```
+
+Reasoning:
+
+- `S-02`, `S-03`, and `S-04` are visually accepted.
+- `S-05` and `S-06` are already in the keep set.
+- `S-07`, `S-08`, and `S-09` continue the early sequence and cover useful variation:
+  - `S-07`: environment pressure against tiny traveler.
+  - `S-08`: body/boot/mud detail.
+  - `S-09`: large bridge/shadow pressure begins.
+
+## Recommended R5 Package Split
+
+Use one clean visible child session.
+
+Package 1:
+
+```text
+S-07
+```
+
+Wait by workflow:
+
+- first check after 10 minutes;
+- if incomplete, poll every 1 minute;
+- at 15 minutes total, inspect status or report timeout/blocker.
+
+After Package 1, parent should check only the compact handoff. The user checks image acceptance.
+
+Package 2, only if Package 1 does not fail the brief-ownership rule:
+
+```text
+S-08
+S-09
+```
+
+Process sequentially inside the same child session. Apply the same per-slide wait discipline and record context growth.
+
+If Package 1 fails brief ownership, stop R5 and repair the workflow contract before generating more images.
+
+## Recommended Task Settings
+
+```text
+running mode: prompt and generate
+review: none
+completion gate: required
+main-session audit: none unless child handoff reports warning / blocked / failed
+child-session capacity target: measure 1 shot, then 2 more shots in the same child session
+overwrite allowed: no
+```
+
+## Required Production Route
+
+For every image-generation prompt:
+
+- Use the `gpt-image-2` Skill Advisor / host-native workflow.
+- Run the Skill mode check:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+- Read the relevant `gpt-image-2` Skill template/reference file before drafting the prompt.
+- Record the Skill mode, template/reference path, and source files actually read.
+- Use `visual-system/style/style-prompt-master.md`.
+- Use `visual-system/characters/character-reference-registry.md` and active character image references when visible identity/body continuity matters.
+- Visible character identity must not be generated from text-only prompts.
+
+## Main-Session Expectations
+
+The main session should:
+
+- create exactly one clean visible child session for R5;
+- send Package 1 only;
+- wait by workflow and read only the compact handoff unless warning/blocked/failed appears;
+- ask the user to inspect the image;
+- continue the same child session with Package 2 only after Package 1 passes the brief-ownership gate and the image is accepted;
+- record observed context usage after each package if the user reports it;
+- avoid opening prompt/image/return details unless the child reports a problem or the user requests inspection.
+
+## Child-Session Expectations
+
+The child session should:
+
+- write visible-child-owned brief and packet before launching each worker;
+- launch one internal one-shot worker per slide/version;
+- keep each worker terminal/closed after return;
+- prevent worker-generated brief/packet;
+- write no review files;
+- perform the expanded completion gate;
+- return compact handoffs.
+
+The compact handoff must explicitly include:
+
+- status;
+- completion gate result;
+- brief owner result for each slide;
+- worker ids;
+- artifact paths;
+- image hashes;
+- review mode/result;
+- context usage if available;
+- whether parent should inspect anything.
+
+## Non-Goals For R5
+
+- Do not invoke CCPE.
+- Do not ask GPT V2 for final prompts.
+- Do not call Minimax MCP or external vision/review tools.
+- Do not reuse old R4 workers.
+- Do not overwrite any slide artifacts.
+- Do not broaden production beyond the R5 package split until the brief-ownership rule passes.
+- Do not treat R4 as a clean mechanism pass.
+
+## Current Archive Updates Already Made
+
+The current session already updated:
+
+```text
+projects/2026-06-23-在路上/project.md
+projects/2026-06-23-在路上/execution-plan.md
+projects/2026-06-23-在路上/slides/slides.md
+projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/task-log.md
+```
+
+These files record:
+
+- `S-02`, `S-03`, and `S-04` images are user accepted.
+- R4 has a brief-ownership mechanism issue.
+- R4 has a `S-04` sequencing warning.
+- Broad production remains paused until the mechanism decision is made.
+
+The user plans to manually commit all current changes after this handoff is written.

projects/2026-06-23-在路上/execution-plan.md

@@ -67,13 +67,13 @@ Round 1 success means these are true:
 - R2 completed: five-shot visual trial generated and accepted by user.
 - R3 character-consistency workflow test completed: `s21-v3`, `s14-v2`, and `s25-v2` generated through the Advisor / visible-reference path and accepted by user. Quality route is usable; duplicate `Agent S-14` remains a historical mechanism defect from this run.
 - R3 third-round workflow test completed: `session-04` generated `s33-v2` and `s06-v2` through a clean user-visible child session plus internal one-shot SubAgents. Both images were accepted by user; review files were absent as requested; no duplicate-worker defect was observed in this two-slide scope.
-- R4 planned: formal-run simulation from a clean committed baseline. Candidate scope is `S-02`, `S-03`, and `S-04`, using one clean child session with up to 3 slides, main-session no-audit default, child-session mandatory completeness gate, and optional review only if requested.
+- R4 completed visually with mechanism warnings: clean child session `019ef8cf-4ab1-7b41-877b-979dc25b0a76` generated `s02-v2`, `s03-v2`, and `s04-v2`; user accepted all three images and review files were absent as requested. Mechanism is not a clean pass because brief/packet files were written by slide Agents rather than prewritten by the visible child session; `S-04` also has a worker-launch sequencing warning.
 
 ## Active Batch
 
-Broad generation trial remains paused until R4 formal-run simulation starts.
+Broad generation trial remains paused until the R4 mechanism decision is made.
 
-No single-shot gate is active. The five-shot visual trial is accepted. The character-consistency workflow test is also visually accepted for S-21, S-14, S-25, S-33, and S-06. The clean `session-04` route is accepted as the basis for R4. R4 should test whether one child session can carry 3 consecutive unfinished shots while the main session only consumes the child handoff.
+No single-shot gate is active. The five-shot visual trial is accepted. The character-consistency workflow test is also visually accepted for S-21, S-14, S-25, S-33, and S-06. R4 images for S-02, S-03, and S-04 are visually accepted, but the workflow needs a mechanism decision before broader production because child-session brief ownership did not match the desired boundary.
 
 Current broad-run issue report:
 
@@ -93,15 +93,11 @@ slides/broad-generation-self-audit-v1.md
 
 ## Next Decision
 
-Run R4 formal-run simulation:
+Resolve R4 formal-run mechanism:
 
-- Recommended scope: `S-02`, `S-03`, `S-04`.
+- R4 scope `S-02`, `S-03`, and `S-04` is visually complete and user accepted.
-- Use one clean user-visible child session with up to 3 slides.
+- Keep the quality route from this test: `gpt-image-2` Advisor prompt/spec, active style master, active character registry when needed, and visible reference images loaded before generation.
-- Main session waits 10 minutes after dispatch, then polls every 1 minute up to 15 minutes total.
+- Decide whether the visible child session must generate `sNN-vN-brief.md` and `sNN-vN-packet.md` before launching a one-shot slide worker. Current `session-05` evidence shows workers generated brief/packet for S-02, S-03, and S-04.
-- Main session reads only the child handoff by default and does not audit prompt / return / ledger / image details when completion gate passes.
+- Keep the duplicate-worker guard: one slide/version must have exactly one owning worker and one prompt/image/return target set.
-- Child session performs mandatory completeness gate. Instruction / visual review is optional and should run only if the task asks for review.
+- Keep main-session production behavior lightweight: wait by workflow, read child handoff by default, and inspect details only on warning/blocked/failed.
-- Start brief/orchestration work as clean user-visible Codex child/session threads and record their thread ids.
+- Future production should not broaden until the brief ownership rule is made explicit, especially for slides without GPT skeleton expansion.
-- Inside a child session, start one internal SubAgent / one-shot slide worker per slide; do not create long-lived reusable slide agents.
-- Add a duplicate-worker guard: one slide/version must have exactly one owning worker and one prompt/image/return target set.
-- Keep the quality route from this test: `gpt-image-2` Advisor prompt/spec, active style master, active character registry, and visible reference images loaded before generation.
-- If R4 succeeds, the project does not need to stay limited to only 3 total new shots; future production can use broader approved scopes or multiple child sessions while preserving the child-session gate.

projects/2026-06-23-在路上/project.md

@@ -60,7 +60,7 @@ The emotional center is not victory. It is continuing.
 
 ## Next Action
 
-Prepare R4 formal-run simulation.
+Inspect R4 formal-run mechanism and decide the next production plan.
 
 ```text
 handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md
@@ -72,6 +72,8 @@ The v2-from-anchor transparent character slice set is accepted. The C repair sam
 
 Third-round mechanism test result: `session-04` used a clean user-visible Codex child thread (`019ef876-c5b2-7202-b204-774cb5fb95fd`) and internal one-shot SubAgents for `Agent S-33` (`019ef87a-3aa1-75a2-a97c-de8f57e234e1`) and `Agent S-06` (`019ef886-8797-7243-9427-d45065fcfc47`). It produced `s33-v2-image.png` and `s06-v2-image.png`; both were accepted by user. Review files were intentionally absent. Prompt traces record `gpt-image-2` Advisor mode plus style and character references.
 
-Fourth-run plan: start from a clean committed baseline and simulate formal production. The main orchestration session should not audit child outputs by default; the child session must return a short handoff with mandatory completeness gate results. Recommended first R4 scope is `S-02`, `S-03`, and `S-04`: three early unfinished shots after accepted `S-01`, using new append-only versions from existing broad-run `v1` evidence.
+Fourth-run result: `session-05` used clean child thread `019ef8cf-4ab1-7b41-877b-979dc25b0a76` for `S-02`, then continued the same child thread for `S-03` and `S-04`. User accepted `s02-v2-image.png`, `s03-v2-image.png`, and `s04-v2-image.png`; review files were intentionally absent. `S-02` completed in about 13m55s and the child context was reported by user as about 107K tokens at that point.
 
-Broad generation remains paused until R4 starts. The accepted quality route is `gpt-image-2` Advisor prompt/spec plus active style master, active character registry, and visible reference images loaded before generation. If R4 succeeds, production does not need to stay limited to only three total shots; future work can dispatch broader approved scopes while preserving child-session completion gates and one-shot internal Agents.
+R4 mechanism finding: image quality route is usable, but the workflow is not a clean mechanism pass. Evidence in worker returns shows `S-02`, `S-03`, and `S-04` brief/packet files were written by their slide Agents rather than prewritten by the visible child session. `S-04` also has a sequencing warning: the worker launched before visible-child brief/packet prewrite confirmation. No overwrite, duplicate worker, or review-file defect was observed in `session-05`.
+
+Broad generation remains paused until the R4 mechanism decision is made. The accepted quality route is `gpt-image-2` Advisor prompt/spec plus active style master, active character registry when needed, and visible reference images loaded before generation. The next production plan should decide whether brief generation belongs to the visible child session before worker launch, especially when no GPT skeleton expansion exists.

projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md

@@ -0,0 +1,87 @@
+# S-02 Slide Brief V2
+
+## Status
+
+- Slide id: `s02`
+- Version: `s02-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Status: ready for prompt; generation blocked unless visible-context references can be loaded by the host session
+
+## Source Basis
+
+- `projects/2026-06-23-在路上/slides/slides.md`: `s02` is Prelude / 0:00-0:16, function "Tiny traveler enters the world along the muddy road", Style Anchor A, broad-run `s02-v1-image.png` exists as historical evidence only.
+- `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`: S-02 is "行者进入世界"; a very small back-view traveler appears in the distance, carrying an old backpack and walking slowly along the muddy road; it transitions from "empty road" to "person on the road".
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`: world before face; road before pose; the traveler is a small coordinate that keeps moving.
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`: use Module A landscape-led muddy road and the global documentary road-film kernel.
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`: visible distant traveler still needs active walking/back image references.
+
+## Slide Function
+
+Introduce the traveler into the already established cold world. S-01 showed the road before the person; S-02 lets a tiny back-view traveler enter that road without becoming the subject of a portrait.
+
+## First Read
+
+Cold muddy road and large landscape first; tiny walking traveler second. The road/world must dominate before the viewer notices the figure.
+
+## Risk Class
+
+- `distant human`
+- Back/walking identity continuity
+- Landscape-led road composition
+
+## Reference Plan
+
+- Style module: A / Landscape-Led Road.
+- Active style reference:
+  - `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+- Style calibration reference:
+  - `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+- Character references:
+  - `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-back-reference-crop-tight.png` as `Image 1`, primary walking-back continuity.
+  - `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-back-v2-from-anchor.png` as `Image 2`, distant walking-away full-body silhouette continuity.
+  - `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-back-reference-crop-tight.png` as `Image 3`, supporting back-view coat/backpack identity.
+
+## Visual Requirements
+
+- 16:9 landscape-led cinematic documentary road-film still.
+- Cold morning / pre-dawn Prelude mood.
+- Long muddy road with wet ruts, frost, rough earth, puddles, stones, sparse terrain, and gray-white distant light.
+- The traveler appears far away as a small back-view or slight back-side figure, walking slowly along the road with an old backpack and long dark road-worn coat.
+- The image must feel like the person has entered a world larger than him, not like a hero entrance.
+
+## Character Handling
+
+Use real image references for back/walking continuity. Do not generate the visible traveler from text-only identity.
+
+The traveler should preserve old backpack, long dark worn coat, shoulder line, practical road-worn silhouette, dark trousers, worn boots, and upright but tired walking posture. No face is needed.
+
+## Composition / Scale
+
+- Traveler is tiny and distant, roughly `1/20` to `1/12` of frame height, or similarly secondary.
+- Place the traveler along the road line in the far lower-middle or lower-third depth.
+- Road/world/horizon dominate most of the frame.
+- The road should pull the eye forward from foreground mud to the distant figure.
+
+## Light / Color / Texture
+
+Cold gray, earth brown, dark khaki, wet blue-gray, coal black, low saturation. Frost and wet mud must be tactile. Avoid postcard sunrise, warm travel mood, clean inspirational road, or heroic backshot.
+
+## Forbidden Drift
+
+- No portrait, no face, no front view, no looking back.
+- No heroic entrance or centered hero stance.
+- No travel-ad scenery, motivational poster, fashion photography, clean road, paved highway, vehicles, readable signs, text, logos, UI, neon, cyberpunk, fantasy portal, or companion figure.
+- No enlarging the traveler beyond a tiny secondary scale.
+
+## Acceptance Checklist
+
+- S-02 reads as "行者进入世界" through a tiny traveler entering the cold muddy road world.
+- Module A road/world dominates.
+- Traveler is small, back-view/walking, and identity-readable through active references.
+- No review file is written.
+- No S-03/S-04 or other slide is touched.
+
+## Prompt Handoff
+
+Use the `gpt-image-2` Skill Advisor / host-native workflow, with Module A and the selected active character references. The final prompt must state that `Image 1`, `Image 2`, and `Image 3` must be loaded into visible context before host generation. If this worker cannot load those references into the generation context, block before generation rather than producing a text-only identity-critical image.

projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md

@@ -0,0 +1,143 @@
+# S-02 Slide Agent Packet V2
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-02`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+
+## Target Slide
+
+- Project: `projects/2026-06-23-在路上`
+- Task: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency`
+- Session: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05`
+- Slide id: `s02`
+- Resolved version: `s02-v2`
+
+## Allowed Actions
+
+- Read this packet and the brief.
+- Read only the listed input files and required `gpt-image-2` Skill files.
+- Write only the S-02 target prompt and return artifacts for `s02-v2`.
+- Generate one image only if the host generation context can load the required real image references as `Image 1`, `Image 2`, and `Image 3`.
+
+Do not process `S-03`, `S-04`, or any slide except `S-02/s02-v2`. Do not create a review file. Do not overwrite any existing target.
+
+## Input Files
+
+- Brief: `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- Slide table: `projects/2026-06-23-在路上/slides/slides.md`
+- S-02 source: `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`
+- Visual system: `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- Style master: `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- Character registry: `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+- R4 handoff: `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+
+## Output Files
+
+- Target brief path: `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- Target packet path: `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+- Session packet path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s02-v2-slide-agent-packet.md`
+- Target prompt path: `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+- Target image path: `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+- Target return path: `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- Session return path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+- Target review path: disabled; do not write `projects/2026-06-23-在路上/slides/s02/s02-v2-review.md`
+
+## Artifact Version / Write Policy
+
+- Version scan result: existing `s02-v*-*` files are limited to `s02-v1-image.png`; `s02-v2` is the next append-only version.
+- Overwrite allowed: no.
+- Before writing prompt, image, review, packet, or return, check whether that path already exists.
+- If any target exists, stop and write only a blocker return if possible; do not overwrite.
+
+## Reference Roles
+
+Load these real image references into visible context before host generation and name their roles in the prompt:
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-back-reference-crop-tight.png`
+  - Role: primary walking-back continuity for backpack placement, coat silhouette, shoulder line, and continued movement.
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-back-v2-from-anchor.png`
+  - Role: distant walking-away full-body silhouette continuity.
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-back-reference-crop-tight.png`
+  - Role: supporting back-view identity for backpack, coat length, shoulder line, and non-hunched road-worn posture.
+
+Active style reference:
+
+- `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: Module A landscape-led muddy-road composition, cold wilderness, road-before-person.
+- `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall style calibration / avoid-drift guard; name for traceability, attach only if useful and available.
+
+## Visual-System Requirements
+
+- Use `visual-system/style/style-prompt-master.md` global style kernel and Module A.
+- Treat S-02 as a distant/back-view visible-traveler shot: character references are mandatory even though the traveler is tiny.
+- Use the S-02 source basis: Prelude, "行者进入世界"; a very small back-view traveler appears far away, carrying an old backpack and walking slowly along the muddy road.
+- The traveler must remain secondary and readable mainly through walking/back silhouette, coat, backpack, and posture.
+
+## Gpt-Image-2 Advisor Rule
+
+Before writing the prompt:
+
+1. Run:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+2. Read:
+
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+Use `concept-scene.md` for road-scene structure and `minimalist-mood-scene.md` for restraint unless a stronger reason for a different scene template is recorded.
+
+## Prompt File Requirements
+
+Write `s02-v2-prompt.md` with:
+
+- mode record;
+- Skill/template/reference paths read;
+- source files actually read;
+- style module A and active style reference;
+- character image references and roles;
+- operation notes;
+- final prompt;
+- negative / avoid;
+- output intent.
+
+The final prompt must be copyable into another image-generation platform and must explicitly state that the real reference images must be loaded as `Image 1`, `Image 2`, and `Image 3`.
+
+## Host Generation Evidence Rule
+
+Use host-native image generation unless mode check says Garden/API is available and required. For current project policy, expected path is:
+
+- Garden/API expected: no.
+- Advisor prompt/spec first.
+- Real references loaded into visible context.
+- Host `image_gen` generation.
+- Copy/save generated final image to `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`.
+
+If this worker cannot load references into visible generation context or cannot save the generated image, block before generation and name the exact missing condition in the return.
+
+## Generation Rule
+
+Generate exactly one final image for `s02-v2` only when required references are loaded. Do not generate from text-only identity for this visible-traveler shot.
+
+## Review Rule
+
+Review disabled. Do not write `s02-v2-review.md` and do not call external vision/review services.
+
+## Return File
+
+Write:
+
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+
+Then stop.

projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md

@@ -0,0 +1,131 @@
+# S-02 V2 Prompt
+
+## Mode Record
+
+Mode check command:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+Result:
+
+```json
+{
+  "mode": "B-or-C",
+  "recommendation": "host-or-advisor",
+  "garden_mode_enabled": false,
+  "has_api_key": false,
+  "base_url": "https://api.openai.com/v1",
+  "model": "gpt-image-2",
+  "env_flag_value": "(unset)",
+  "summary": "MODE B / C · 未启用 Garden。若宿主 Agent 自带图像工具（image_generation / dalle / mcp__*image* 等）→ MODE B：把 prompt 交给宿主出图。若宿主无图像工具 → MODE C：仅产出高质量 prompt 给用户。"
+}
+```
+
+Interpretation for this task:
+
+- Use `gpt-image-2` as Advisor / Host-Native prompt workflow.
+- Do not call Garden/API generation scripts.
+- Use the host image tool only with the three visible character references loaded as `Image 1`, `Image 2`, and `Image 3`.
+
+## Skill Files Read
+
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+Template adaptation:
+
+- `concept-scene.md` supplies foreground / midground / background / lighting structure for the long-road scene.
+- `minimalist-mood-scene.md` supplies restraint: tiny subject, low element count, subdued palette, no heroic symbolism, no decorative excess.
+
+## Source Files Actually Read
+
+- `AGENTS.md`
+- `docs/workflows/slide-task-orchestration.md`
+- `docs/workflows/slide-task-templates.md`
+- `projects/2026-06-23-在路上/project.md`
+- `projects/2026-06-23-在路上/execution-plan.md`
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- `projects/2026-06-23-在路上/slides/slides.md`
+- `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+- `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+
+## Reference Inputs
+
+- Style module: `A / Landscape-Led Road`
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: landscape-led muddy-road composition, cold wilderness, road-before-person.
+- Style calibration reference: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall A/B/C/D continuity and avoid-drift guard; named for traceability, not required as an image attachment for this generation.
+- Character reference `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-back-reference-crop-tight.png`
+  - Role: primary walking-back continuity for backpack placement, long dark coat silhouette, shoulder line, and continued movement.
+- Character reference `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-back-v2-from-anchor.png`
+  - Role: distant full-body walking-away silhouette continuity.
+- Character reference `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-back-reference-crop-tight.png`
+  - Role: supporting back-view identity for backpack, coat length, shoulder line, and non-hunched road-worn posture.
+
+## Operation Notes
+
+- Running mode: `prompt and generate`.
+- Review: disabled; do not write `s02-v2-review.md`.
+- Overwrite allowed: no.
+- Version scan confirmed only `s02-v1-image.png` exists before this run; `s02-v2` is the next append-only version.
+- First read order: cold muddy road and large Prelude landscape first; tiny distant walking traveler second.
+- Character handling: the visible traveler is tiny and secondary, but identity must be guided by real back/walking references, not text-only character invention.
+- Scale target: traveler roughly `1/20` to `1/12` of frame height, or similarly secondary.
+- Output aspect: 16:9 landscape cinematic still.
+
+## Final Prompt
+
+Create a 16:9 photorealistic cinematic documentary road-film still for the MV shot S-02, Prelude, "行者进入世界".
+
+The real reference images must be loaded and used exactly as:
+
+- Image 1: primary walking-back continuity reference. Preserve the walking-away backpack placement, long dark road-worn coat silhouette, shoulder line, and continued movement.
+- Image 2: distant walking-away full-body silhouette reference. Use it to keep the tiny figure readable as the same traveler even at long distance.
+- Image 3: supporting back-view identity reference. Preserve the backpack shape, coat length, shoulder line, and upright non-hunched but weighted posture.
+
+Scene / world:
+
+A cold pre-dawn or cold early-morning wilderness road scene, Module A landscape-led muddy road. The frame is dominated by wet rutted earth, frost, rough stones, sparse terrain, gray-white distant light, shallow puddles, and a long muddy road that recedes into the cold distance. The world is larger than the person; the road is longer than the person. The road and landscape must be the first read.
+
+Narrative moment:
+
+After an empty road opening, a tiny traveler finally enters the world. He appears far away as a small back-view walking figure, carrying an old backpack, moving slowly along the muddy road. This is not a hero entrance; it is the first quiet proof that someone is on the road.
+
+Composition:
+
+Wide landscape-led composition. Use foreground mud, frost, puddles, stones, and road ruts to pull the eye toward the distant traveler. Place the traveler small and secondary along the road line in the far lower-middle or lower third depth, roughly 1/20 to 1/12 of the frame height. Road, mud, cold air, low horizon, and gray wilderness dominate most of the image. The viewer should first feel the road and world, then notice the tiny moving figure.
+
+Traveler:
+
+Use Images 1, 2, and 3 to preserve the same road-worn back-view identity: old dark backpack, long dark worn coat, practical non-fashion silhouette, dark trousers, worn boots, tousled dark hair seen from behind, upright but tired walking posture. He is tiny, quiet, and readable mainly through coat, backpack, shoulder line, walking/back posture, and scale. No face, no front view, no looking back.
+
+Lighting / mood:
+
+Cold gray, earth brown, dark khaki, wet blue-gray, coal black, low saturation. Pre-dawn haze or cold morning thin light. Tactile wet mud, frost in the ruts, shallow puddles, rough ground. Quiet continuation, restrained poetic realism, documentary texture.
+
+Style:
+
+Photorealistic cinematic documentary road-film still. Real documentary texture with restrained poetic realism. Environment first, road before person, no victory, no tourism, no fashion, no poster heroism.
+
+## Negative / Avoid
+
+Avoid portrait framing, face visibility, front view, looking back, heroic centered stance, fashion photography, clean commercial clothing, motivational poster mood, travel advertising, scenic tourism, beautiful postcard sunrise, broad sunset, paved highway, clean road, vehicles, readable signs, text, labels, logos, watermarks, UI, neon, cyberpunk, fantasy portal, magical glow, crowd, companion figure, camping mood, warm inspirational road, or enlarging the traveler beyond tiny secondary scale.
+
+## Output Intent
+
+Generate one final PNG image for:
+
+```text
+projects/2026-06-23-在路上/slides/s02/s02-v2-image.png
+```
+
+The image should satisfy the S-02 function: transition from the empty road into "person on the road" while preserving the project's core rule that the world and road come before the traveler. The actual reference images must be loaded into visible host generation context before image generation; a local path written in this prompt is traceability only and does not count as loading the reference.

projects/2026-06-23-在路上/slides/s02/s02-v2-return.md

@@ -0,0 +1,121 @@
+# S-02 V2 Slide Agent Return
+
+## Status
+
+- Worker: `Agent S-02`
+- Slide: `s02`
+- Version: `s02-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: blocked
+- Blockers: host generation reference-loading unavailable in this worker
+
+## Files Written
+
+- Brief path: `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- Packet path: `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+- Session packet path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s02-v2-slide-agent-packet.md`
+- Prompt path: `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+- Image path: not written; `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png` is absent
+- Review path: not written; review disabled
+- Return path: `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- Session return path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+
+## Completeness Gate
+
+- Version resolved by scan: pass. Existing `s02-v*-*` files before writing showed only `s02-v1-image.png`, so `s02-v2` was the next append-only version.
+- No overwrite: pass. Planned target text artifacts were missing before write.
+- Brief present: pass.
+- Packet present: pass.
+- Prompt present: pass.
+- Image present: blocked. Required image file is absent because generation was not executed.
+- Review absent as instructed: pass. No `s02-v2-review.md` was written.
+- Owning worker: pass. This return represents one one-shot worker for `S-02/s02-v2`.
+- Reference loading: blocked. The final prompt requires `Image 1`, `Image 2`, and `Image 3` to be loaded into visible host generation context, but this worker has no available interface to attach those local reference PNGs to the host image generation call.
+
+## Advisor Result
+
+Mode check command:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+Mode result:
+
+```json
+{
+  "mode": "B-or-C",
+  "recommendation": "host-or-advisor",
+  "garden_mode_enabled": false,
+  "has_api_key": false,
+  "base_url": "https://api.openai.com/v1",
+  "model": "gpt-image-2",
+  "env_flag_value": "(unset)"
+}
+```
+
+Interpretation used:
+
+- Garden/API generation expected: no.
+- Advisor / host-native prompt workflow used: yes.
+- Prompt/spec was written first through the `gpt-image-2` Advisor workflow.
+
+## Generation Result
+
+- Host image generation completed: no.
+- Reason: this visible-traveler shot requires active character references loaded into the generation context. The available worker tool interface does not provide a way to attach the required local reference images as `Image 1`, `Image 2`, and `Image 3`.
+- Safety action: blocked before text-only identity-critical generation.
+
+## Host Generation Evidence
+
+- Garden/API expected: no.
+- Host tool call made: no.
+- Host generated-image source path: none.
+- Copied project asset: none.
+- Image hash: none.
+
+## Review Result
+
+- Review mode: none / disabled by task.
+- Review file written: no.
+- External vision/review services called: no.
+
+## Reference Roles
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-back-reference-crop-tight.png`
+  - Role: primary walking-back continuity for backpack placement, long dark coat silhouette, shoulder line, and continued movement.
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-back-v2-from-anchor.png`
+  - Role: distant full-body walking-away silhouette continuity.
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-back-reference-crop-tight.png`
+  - Role: supporting back-view identity for backpack, coat length, shoulder line, and non-hunched road-worn posture.
+- Active style reference named in prompt: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: Module A landscape-led muddy-road composition, cold wilderness, road-before-person.
+- Style calibration reference named in prompt: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall style calibration / avoid-drift guard.
+
+## Blockers
+
+Exact missing condition:
+
+```text
+The required local character reference images cannot be loaded into this worker's host image-generation context as Image 1 / Image 2 / Image 3, and visible-character identity cannot be generated from a text-only prompt under the project rules.
+```
+
+The visible parent session should load the three listed character reference images into conversation/image-generation context, run the saved prompt in `s02-v2-prompt.md`, and save/copy the selected output to:
+
+```text
+projects/2026-06-23-在路上/slides/s02/s02-v2-image.png
+```
+
+## Parent Summary
+
+`S-02/s02-v2` brief, packet, and Advisor/host-native prompt are written. Generation is blocked because this worker cannot attach the mandatory character references to host image generation. No image and no review file were written. No S-03/S-04 work was performed.
+
+## Next Action
+
+Parent/visible session should perform reference-capable host generation for `s02-v2` using the saved prompt and required `Image 1` / `Image 2` / `Image 3` roles, then place the PNG at `slides/s02/s02-v2-image.png`.
+
+## Agent Lifecycle
+
+This slide worker is terminal / closed after this return. Do not continue or reuse this worker for repair, retry, S-03, S-04, or any other slide.

projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md

@@ -0,0 +1,87 @@
+# S-03 Slide Brief V2
+
+## Status
+
+- Slide id: `s03`
+- Version: `s03-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Status: ready for prompt; generation blocked unless visible-context references can be loaded by the host session
+
+## Source Basis
+
+- `projects/2026-06-23-在路上/slides/slides.md`: `s03` is Prelude / 0:00-0:16, function "Old boot and backpack detail connects body to road", style anchor `character detail`, broad-run `s03-v1-image.png` exists as historical evidence only.
+- `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`: S-03 is "行囊与路的关系"; low-position local shot, old boot steps over the wet mud edge, backpack lower edge sways slightly, not a shoe advertisement.
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`: old boots and old backpack are identity motifs; the body connects to mud, road, and continuation.
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`: use global documentary road-film kernel and the `Character detail with road texture` mapping, controlled mainly by Module A.
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`: object/body detail should use the master anchor or relevant crop/slice when coat, backpack, boots, or posture identity matters.
+
+## Slide Function
+
+Turn the traveler's body and props into evidence of the road. S-03 should not introduce face or hero identity; it should make the old boot, wet mud, backpack lower edge, and walking body detail feel inseparable from the road.
+
+## First Read
+
+Wet mud, boot contact, and road texture first; backpack / coat movement second. The viewer should feel bodily cost and continuation before reading the image as a person shot.
+
+## Risk Class
+
+- `object/body detail`
+- Prop continuity: old boot, backpack, long dark coat
+- Low-angle road texture
+
+## Reference Plan
+
+- Style module: A / Landscape-Led Road, adapted for close body / prop detail.
+- Active style reference:
+  - `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+- Style calibration reference:
+  - `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+- Character / prop references:
+  - `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-codex-posture-test.png` as `Image 1`, master identity / clothing / backpack / boots / posture reference.
+  - `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-side-reference-crop-tight.png` as `Image 2`, side walking posture, coat motion, backpack placement, stride, and boot reference.
+  - `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-side-v2-from-anchor.png` as `Image 3`, full-body walking-side silhouette and boot / coat / backpack continuity.
+
+## Visual Requirements
+
+- 16:9 photorealistic cinematic documentary road-film still.
+- Low camera near wet mud, boot, road edge, and lower body.
+- Old boot steps through or just past the wet muddy edge; mud compresses under the sole and small wet grit gathers around the boot.
+- The lower edge of the old backpack or backpack strap is visible higher in frame, lightly swinging with the walking motion.
+- Coat hem and dark road-worn trouser / boot relationship should be readable without showing face.
+- The image must feel like body meeting road, not a product shot of footwear.
+
+## Character Handling
+
+Use real active references for prop and body continuity. No face is needed and no face should appear. Do not invent new stylish boots, a new backpack, or a fashion silhouette from text-only prompting.
+
+## Composition / Scale
+
+- Low-position local detail, 16:9 horizontal.
+- Boot / mud interaction is the first-read anchor, but the boot should not be polished or centered like an advertisement.
+- Use shallow foreground mud texture, road ruts, pebbles, puddle edge, and moving leg / coat / backpack fragments.
+- Keep the body cropped naturally: lower leg, boot, coat hem, backpack lower edge or strap. No full portrait.
+
+## Light / Color / Texture
+
+Cold gray, earth brown, dark khaki, wet blue-gray, coal black, low saturation. Morning / pre-dawn damp cold. Mud must be tactile: wet grit, pressure, soft splash, rough stones, frost residue, road wear.
+
+## Forbidden Drift
+
+- No shoe advertisement, product hero lighting, clean boot close-up, fashion styling, luxury outdoor campaign, crisp studio product shot, readable brand logo, or polished gear catalog.
+- No face, no front portrait, no heroic full-body pose, no looking back.
+- No clean road, paved highway, vehicle, city UI, neon, cyberpunk, fantasy glow, motivational poster, broad sunrise, readable text, labels, logos, or watermark.
+- Do not process S-02, S-04, or any other slide.
+
+## Acceptance Checklist
+
+- S-03 reads as "行囊与路的关系": old boot, wet mud, backpack / coat movement, continuing body.
+- The first impression is bodily contact with road texture, not a shoe ad.
+- Active character / prop references are selected and named.
+- Module A road texture and global documentary road-film style are used.
+- No review file is written.
+- No S-02/S-04 or other slide is touched.
+
+## Prompt Handoff
+
+Use the `gpt-image-2` Skill Advisor / host-native workflow, with Module A adapted for body / prop detail and the selected active character references. The final prompt must state that `Image 1`, `Image 2`, and `Image 3` must be loaded into visible context before host generation. If this worker cannot load those references into the generation context, block before generation rather than producing a text-only prop / identity-critical image.

projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md

@@ -0,0 +1,143 @@
+# S-03 Slide Agent Packet V2
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-03`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+
+## Target Slide
+
+- Project: `projects/2026-06-23-在路上`
+- Task: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency`
+- Session: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05`
+- Slide id: `s03`
+- Resolved version: `s03-v2`
+
+## Allowed Actions
+
+- Read this packet and the brief.
+- Read only the listed input files and required `gpt-image-2` Skill files.
+- Write only the S-03 target prompt and return artifacts for `s03-v2`.
+- Generate one image only if the host generation context can load the required real image references as `Image 1`, `Image 2`, and `Image 3`.
+
+Do not process `S-02`, `S-04`, or any slide except `S-03/s03-v2`. Do not create a review file. Do not overwrite any existing target.
+
+## Input Files
+
+- Brief: `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- Slide table: `projects/2026-06-23-在路上/slides/slides.md`
+- S-03 source: `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`
+- Visual system: `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- Style master: `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- Character registry: `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+- R4 handoff: `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+
+## Output Files
+
+- Target brief path: `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- Target packet path: `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- Session packet path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s03-v2-slide-agent-packet.md`
+- Target prompt path: `projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md`
+- Target image path: `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png`
+- Target return path: `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- Session return path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+- Target review path: disabled; do not write `projects/2026-06-23-在路上/slides/s03/s03-v2-review.md`
+
+## Artifact Version / Write Policy
+
+- Version scan result: existing `s03-v*-*` files are limited to `s03-v1-image.png`; `s03-v2` is the next append-only version.
+- Overwrite allowed: no.
+- Before writing prompt, image, review, packet, or return, check whether that path already exists.
+- If any target exists, stop and write only a blocker return if possible; do not overwrite.
+
+## Reference Roles
+
+Load these real image references into visible context before host generation and name their roles in the prompt:
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-codex-posture-test.png`
+  - Role: master continuity for clothing, old backpack, old boots, body posture, and prop relationship.
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-side-reference-crop-tight.png`
+  - Role: side walking posture, coat movement, backpack placement, stride, and boot behavior.
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-side-v2-from-anchor.png`
+  - Role: full-body walking-side silhouette continuity for boot / coat / backpack scale.
+
+Active style reference:
+
+- `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: Module A muddy-road texture, cold wilderness, road-before-person.
+- `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall style calibration / avoid-drift guard; name for traceability, attach only if useful and available.
+
+## Visual-System Requirements
+
+- Use `visual-system/style/style-prompt-master.md` global style kernel and Module A adapted for character detail with road texture.
+- Treat S-03 as an object/body detail shot: old boot, wet mud, lower body, backpack / coat motion, and continuation.
+- Use the S-03 source basis: low-position local shot, old boot steps over wet mud edge, backpack lower edge sways slightly, not a shoe advertisement.
+- No face is required or allowed.
+
+## Gpt-Image-2 Advisor Rule
+
+Before writing the prompt:
+
+1. Run:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+2. Read:
+
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+Use `concept-scene.md` for foreground / midground / background / lighting structure and `minimalist-mood-scene.md` for restraint unless a stronger reason for a different scene template is recorded.
+
+## Prompt File Requirements
+
+Write `s03-v2-prompt.md` with:
+
+- mode record;
+- Skill/template/reference paths read;
+- source files actually read;
+- style module A and active style reference;
+- character / prop image references and roles;
+- operation notes;
+- final prompt;
+- negative / avoid;
+- output intent.
+
+The final prompt must be copyable into another image-generation platform and must explicitly state that the real reference images must be loaded as `Image 1`, `Image 2`, and `Image 3`.
+
+## Host Generation Evidence Rule
+
+Use host-native image generation unless mode check says Garden/API is available and required. For current project policy, expected path is:
+
+- Garden/API expected: no.
+- Advisor prompt/spec first.
+- Real references loaded into visible context.
+- Host `image_gen` generation.
+- Copy/save generated final image to `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png`.
+
+If this worker cannot load references into visible generation context or cannot save the generated image, block before generation and name the exact missing condition in the return.
+
+## Generation Rule
+
+Generate exactly one final image for `s03-v2` only when required references are loaded. Do not generate a text-only prop / identity-critical body-detail image.
+
+## Review Rule
+
+Review disabled. Do not write `s03-v2-review.md` and do not call external vision/review services.
+
+## Return File
+
+Write:
+
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+
+Then stop.

projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md

@@ -0,0 +1,131 @@
+# S-03 V2 Prompt
+
+## Mode Record
+
+Mode check command:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+Result:
+
+```json
+{
+  "mode": "B-or-C",
+  "recommendation": "host-or-advisor",
+  "garden_mode_enabled": false,
+  "has_api_key": false,
+  "base_url": "https://api.openai.com/v1",
+  "model": "gpt-image-2",
+  "env_flag_value": "(unset)",
+  "summary": "MODE B / C · 未启用 Garden。若宿主 Agent 自带图像工具（image_generation / dalle / mcp__*image* 等）→ MODE B：把 prompt 交给宿主出图。若宿主无图像工具 → MODE C：仅产出高质量 prompt 给用户。"
+}
+```
+
+Interpretation for this task:
+
+- Use `gpt-image-2` as Advisor / Host-Native prompt workflow.
+- Do not call Garden/API generation scripts.
+- Use the host image tool only with the three visible character / prop references loaded as `Image 1`, `Image 2`, and `Image 3`.
+
+## Skill Files Read
+
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+Template adaptation:
+
+- `concept-scene.md` supplies foreground / midground / background / lighting structure for the low road-detail scene.
+- `minimalist-mood-scene.md` supplies restraint: few elements, subdued palette, no product-display heroism, no decorative excess.
+
+## Source Files Actually Read
+
+- `AGENTS.md`
+- `docs/workflows/slide-task-orchestration.md`
+- `docs/workflows/slide-task-templates.md`
+- `projects/2026-06-23-在路上/project.md`
+- `projects/2026-06-23-在路上/execution-plan.md`
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- `projects/2026-06-23-在路上/slides/slides.md`
+- `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+- `projects/2026-06-23-在路上/intake/阶段五-镜头清单骨架.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-*` files were read only as session-05 formatting precedent.
+
+## Reference Inputs
+
+- Style module: `A / Landscape-Led Road`, adapted for `character detail with road texture`
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: muddy-road texture, cold Prelude world, road-before-person continuity.
+- Style calibration reference: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall A/B/C/D continuity and avoid-drift guard; named for traceability, not required as an image attachment for this generation.
+- Character / prop reference `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-codex-posture-test.png`
+  - Role: master continuity for the old backpack, long dark coat, old boots, road-worn body posture, and prop relationship.
+- Character / prop reference `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-side-reference-crop-tight.png`
+  - Role: side walking posture, coat motion, backpack placement, stride, and boot behavior.
+- Character / prop reference `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-side-v2-from-anchor.png`
+  - Role: full-body walking-side silhouette continuity for boot / coat / backpack scale.
+
+## Operation Notes
+
+- Running mode: `prompt and generate`.
+- Review: disabled; do not write `s03-v2-review.md`.
+- Overwrite allowed: no.
+- Version scan confirmed only `s03-v1-image.png` exists before this run; `s03-v2` is the next append-only version.
+- First read order: wet mud and boot contact first; backpack / coat motion second; no face.
+- Character handling: use active references for old boot, backpack, coat, stride, and road-worn posture; do not invent a new fashion / outdoor-ad prop set.
+- Output aspect: 16:9 landscape cinematic still.
+
+## Final Prompt
+
+Create a 16:9 photorealistic cinematic documentary road-film still for the MV shot S-03, Prelude tail, "行囊与路的关系".
+
+The real reference images must be loaded and used exactly as:
+
+- Image 1: master character / prop continuity reference. Preserve the old backpack, long dark road-worn coat, old boots, practical middle-aged traveler posture, and non-fashion silhouette.
+- Image 2: side walking continuity reference. Preserve the walking stride, coat hem motion, backpack placement, and boot behavior.
+- Image 3: full-body walking-side silhouette reference. Use it to keep the cropped lower-body / prop scale consistent with the same traveler.
+
+Scene / world:
+
+A cold pre-dawn or cold early-morning muddy road detail, Module A adapted for body and prop texture. Low camera near the ground. Wet mud, road ruts, grit, pebbles, shallow puddle edge, frost residue, and rough earth dominate the frame. This is the same road-world as the Prelude: documentary, cold, low-saturation, road before person.
+
+Narrative moment:
+
+The body meets the road. An old boot steps across the wet mud edge, compressing the mud under the sole, with a slight drag or small wet splash. Above and behind it, the lower edge of the old backpack or a backpack strap sways lightly with the walking motion. A dark coat hem and lower leg pass through the frame. The shot plants the object anchor for "斑驳行囊" without becoming a gear advertisement.
+
+Composition:
+
+Low-position local detail in a 16:9 horizontal frame. Foreground mud and boot contact are the first read. Midground shows the lower leg, coat hem, and a partial old backpack lower edge or strap in motion. Background remains a blurred cold muddy road with ruts and gray-white morning air. Keep the crop natural and cinematic: no full face, no full portrait, no product-display centering. The boot and backpack are part of a walking body, not isolated products.
+
+Traveler / props:
+
+Use Images 1, 2, and 3 to preserve the same traveler's old dark backpack, long dark worn coat, practical road-worn trousers, old boots, tired but capable walking posture, and non-fashion silhouette. No face is visible. No front portrait. The props look used by life and road, not styled for advertising.
+
+Lighting / mood:
+
+Cold gray, earth brown, dark khaki, wet blue-gray, coal black, low saturation. Soft pre-dawn or cold morning light, damp air, tactile wet mud, rough stones, and frost residue. Quiet bodily persistence. Real documentary texture with restrained poetic realism.
+
+Style:
+
+Photorealistic cinematic documentary road-film still. Real documentary texture with restrained poetic realism. Environment and road texture first, body and props second. No victory, no tourism, no fashion, no poster heroism.
+
+## Negative / Avoid
+
+Avoid shoe advertisement, boot product shot, outdoor gear catalog, clean commercial footwear, crisp studio lighting, fashion photography, luxury travel campaign, readable brand logo, clean road, paved highway, heroic pose, full-body hero framing, front portrait, face visibility, looking back, motivational poster mood, scenic tourism, beautiful postcard sunrise, broad sunset, vehicles, readable signs, text, labels, logos, watermarks, UI, neon, cyberpunk, fantasy portal, magical glow, crowd, companion figure, camping mood, or enlarging the prop into a polished product hero.
+
+## Output Intent
+
+Generate one final PNG image for:
+
+```text
+projects/2026-06-23-在路上/slides/s03/s03-v2-image.png
+```
+
+The image should satisfy the S-03 function: old boot, wet mud, and backpack / coat movement connect the traveler's body to the road while preserving the project's core rule that the road and world come before face or pose. The actual reference images must be loaded into visible host generation context before image generation; a local path written in this prompt is traceability only and does not count as loading the reference.

projects/2026-06-23-在路上/slides/s03/s03-v2-return.md

@@ -0,0 +1,121 @@
+# S-03 V2 Slide Agent Return
+
+## Status
+
+- Worker: `Agent S-03`
+- Slide: `s03`
+- Version: `s03-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: blocked
+- Blockers: host generation reference-loading unavailable in this worker
+
+## Files Written
+
+- Brief path: `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- Packet path: `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- Session packet path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s03-v2-slide-agent-packet.md`
+- Prompt path: `projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md`
+- Image path: not written; `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png` is absent
+- Review path: not written; review disabled
+- Return path: `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- Session return path: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+
+## Completeness Gate
+
+- Version resolved by scan: pass. Existing `s03-v*-*` files before writing showed only `s03-v1-image.png`, so `s03-v2` was the next append-only version.
+- No overwrite: pass. Planned target text artifacts were missing before write.
+- Brief present: pass.
+- Packet present: pass.
+- Prompt present: pass.
+- Image present: blocked. Required image file is absent because generation was not executed.
+- Review absent as instructed: pass. No `s03-v2-review.md` was written.
+- Owning worker: pass. This return represents one one-shot worker for `S-03/s03-v2`.
+- Reference loading: blocked. The final prompt requires `Image 1`, `Image 2`, and `Image 3` to be loaded into visible host generation context, but this worker has no available interface to attach those local reference PNGs to the host image generation call.
+
+## Advisor Result
+
+Mode check command:
+
+```powershell
+node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json
+```
+
+Mode result:
+
+```json
+{
+  "mode": "B-or-C",
+  "recommendation": "host-or-advisor",
+  "garden_mode_enabled": false,
+  "has_api_key": false,
+  "base_url": "https://api.openai.com/v1",
+  "model": "gpt-image-2",
+  "env_flag_value": "(unset)"
+}
+```
+
+Interpretation used:
+
+- Garden/API generation expected: no.
+- Advisor / host-native prompt workflow used: yes.
+- Prompt/spec was written first through the `gpt-image-2` Advisor workflow.
+
+## Generation Result
+
+- Host image generation completed: no.
+- Reason: this body / prop detail shot requires active character / prop references loaded into the generation context. The available worker tool interface does not provide a way to attach the required local reference images as `Image 1`, `Image 2`, and `Image 3`.
+- Safety action: blocked before text-only prop / identity-critical generation.
+
+## Host Generation Evidence
+
+- Garden/API expected: no.
+- Host tool call made: no.
+- Host generated-image source path: none.
+- Copied project asset: none.
+- Image hash: none.
+
+## Review Result
+
+- Review mode: none / disabled by task.
+- Review file written: no.
+- External vision/review services called: no.
+
+## Reference Roles
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-codex-posture-test.png`
+  - Role: master continuity for clothing, old backpack, old boots, body posture, and prop relationship.
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-side-reference-crop-tight.png`
+  - Role: side walking posture, coat movement, backpack placement, stride, and boot behavior.
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-side-v2-from-anchor.png`
+  - Role: full-body walking-side silhouette continuity for boot / coat / backpack scale.
+- Active style reference named in prompt: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+  - Role: Module A muddy-road texture, cold wilderness, road-before-person continuity.
+- Style calibration reference named in prompt: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+  - Role: overall style calibration / avoid-drift guard.
+
+## Blockers
+
+Exact missing condition:
+
+```text
+The required local character / prop reference images cannot be loaded into this worker's host image-generation context as Image 1 / Image 2 / Image 3, and this body / prop detail shot should not be generated from a text-only prompt under the project rules.
+```
+
+The visible parent session should load the three listed character / prop reference images into conversation/image-generation context, run the saved prompt in `s03-v2-prompt.md`, and place the selected output at:
+
+```text
+projects/2026-06-23-在路上/slides/s03/s03-v2-image.png
+```
+
+## Parent Summary
+
+`S-03/s03-v2` brief, packet, and Advisor/host-native prompt are written. Generation is blocked because this worker cannot attach the mandatory character / prop references to host image generation. No image and no review file were written. No S-02/S-04 work was performed.
+
+## Next Action
+
+Parent/visible session should perform reference-capable host generation for `s03-v2` using the saved prompt and required `Image 1` / `Image 2` / `Image 3` roles, then place the PNG at `slides/s03/s03-v2-image.png`.
+
+## Agent Lifecycle
+
+This slide worker is terminal / closed after this return. Do not continue or reuse this worker for repair, retry, S-02, S-04, or any other slide.

projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md

@@ -0,0 +1,100 @@
+# S-04 Slide Brief V2
+
+## Status
+
+- Brief status: ready
+- Worker profile: `video-slide-production`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-04` / `s04-v2`
+- Overwrite allowed: no
+
+## Source Basis
+
+- `AGENTS.md`
+- `docs/workflows/slide-task-orchestration.md`
+- `docs/workflows/slide-task-templates.md`
+- `projects/2026-06-23-在路上/project.md`
+- `projects/2026-06-23-在路上/execution-plan.md`
+- `projects/2026-06-23-在路上/slides/slides.md`
+- `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+
+## Slide Function
+
+`S-04` is the early Section A road shot: cold morning frost and a long road. Its job is to extend the opening world-before-person grammar after `S-01`, keeping the road and cold terrain as the subject.
+
+## First Read
+
+The viewer should first read:
+
+1. cold road and frosted mud;
+2. long worn road receding through a large empty wilderness;
+3. weak pre-dawn / morning haze;
+4. only then, optionally, a tiny traveler far away as a non-identity-critical coordinate.
+
+## Risk Class
+
+- Risk class: `environment-only / distant human optional`
+- Identity critical: no
+- Face requirement: none
+- Character references: not required unless a visible traveler becomes larger than a tiny distant coordinate.
+
+## Reference Plan
+
+- Style module: A / Landscape-Led Road
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`
+- Common calibration reference: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`
+- Character reference: none required for this prompt; if a traveler appears, keep the figure tiny, back/side, and non-face.
+- Prior sequence context: `S-01` accepted opening road grammar, but do not copy its image directly.
+
+## Visual Requirements
+
+- 16:9 horizontal cinematic documentary still.
+- Landscape-led road-film composition.
+- Frost, wet mud, ruts, stones, puddles, sparse terrain.
+- Low saturation: cold gray, earth brown, dark khaki, wet blue-gray.
+- The road must be longer than the person and should dominate the frame.
+- Any traveler must be tiny and secondary, not a portrait or hero backshot.
+
+## Character Handling
+
+No visible identity handling is required. If the image model introduces a distant traveler, the figure should be about `1/20` to `1/12` of the frame, seen from back or side, without readable face detail. Do not attach active character image references for this landscape-led no-face shot unless the parent session deliberately changes the shot to a visible traveler shot.
+
+## Composition / Scale
+
+- Camera: wide road-film frame, slightly low or eye-level observational viewpoint.
+- Foreground: frosted mud, rough ruts, wet stones, dull puddle edges.
+- Midground: long road bending or tapering into cold haze.
+- Background: empty sparse wilderness, low horizon, weak morning haze.
+- Person scale: absent preferred; tiny optional only.
+
+## Light / Color / Texture
+
+- Light: cold early morning / pre-dawn silver-gray, overcast or thin haze.
+- Texture: documentary realism, real mud and frost, no polished scenic gloss.
+- Warmth: almost none; if present, only a weak pale edge in the sky, not sunrise.
+
+## Forbidden Drift
+
+- No travel advertising.
+- No postcard sunrise.
+- No clean inspirational road.
+- No centered heroic traveler.
+- No portrait MV.
+- No vehicle, animal, tent, camp, city, UI, logo, readable text, watermark, fantasy glow, or high-saturation scenery.
+
+## Acceptance Checklist
+
+- First read is cold road / frost / mud, not person.
+- The road feels long and unresolved.
+- Landscape and road dominate the frame.
+- Any traveler is tiny, non-face, and non-identity-critical.
+- Style follows active Module A and global documentary road-film kernel.
+- Review file remains absent because review mode is disabled.
+
+## Prompt Handoff
+
+Build `s04-v2-prompt.md` through `gpt-image-2` Advisor / host-native workflow using Module A. Record the mode check, template/reference files read, active style references, and source files actually read. Do not require character image attachment. If host image generation cannot be accessed or the generated image cannot be saved to `s04-v2-image.png`, return blocked with the exact missing condition.

projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md

@@ -0,0 +1,126 @@
+# S-04 Slide Agent Packet V2
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-04`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-04` / `s04-v2`
+- Overwrite allowed: no
+
+## Target Slide
+
+- Slide: `S-04`
+- Version: `s04-v2`
+- Function: cold morning frost and long road.
+- Style anchor: A / Landscape-Led Road.
+- Risk class: `environment-only / distant human optional`
+
+## Allowed Actions
+
+- Read only required project, workflow, style, character, and prompt-template sources.
+- Write only the `S-04/s04-v2` artifact set and mirrored `session-05` packet/return files.
+- Use `gpt-image-2` as Advisor / host-native prompt workflow.
+- Generate only if host generation can be completed and the PNG can be saved at the target path.
+
+## Input Files
+
+- `AGENTS.md`
+- `docs/workflows/slide-task-orchestration.md`
+- `docs/workflows/slide-task-templates.md`
+- `projects/2026-06-23-在路上/project.md`
+- `projects/2026-06-23-在路上/execution-plan.md`
+- `projects/2026-06-23-在路上/slides/slides.md`
+- `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+## Output Files
+
+- Brief: `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- Packet: `projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md`
+- Prompt: `projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md`
+- Image: `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- Return: `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- Session packet: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s04-v2-slide-agent-packet.md`
+- Session return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`
+- Review: disabled; do not write `s04-v2-review.md`
+
+## Artifact Version / Write Policy
+
+- Existing scan result: `projects/2026-06-23-在路上/slides/s04/` contained only `s04-v1-image.png` before this write.
+- Resolved version: `s04-v2`.
+- Target collision check: all `s04-v2-*` targets were absent before writing.
+- Overwrite allowed: no.
+- Stop on any later target collision.
+
+## Reference Roles
+
+- Style module: A / Landscape-Led Road.
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`.
+- Common style calibration: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`.
+- Character references: none required; this is a no-face, landscape-led road shot. If a traveler appears, keep the figure tiny and non-identity-critical.
+
+## Visual-System Requirements
+
+- Use active `visual-system/style/style-prompt-master.md`.
+- Preserve the global road-film kernel: real documentary texture + restrained poetic realism + landscape-led road-film composition.
+- Use Module A wording for frost, mud, long road, cold dawn / pre-dawn haze, and world-before-person composition.
+- Do not convert the shot into a portrait, heroic backshot, tourism landscape, or clean motivational road image.
+
+## Gpt-Image-2 Advisor Rule
+
+- Run `node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json` before writing the prompt.
+- Use the skill as Advisor / host-native prompt workflow.
+- Record the mode result and skill reference paths in `s04-v2-prompt.md`.
+- Do not call Garden/API generation scripts unless project policy changes.
+
+## Host Generation Evidence Rule
+
+If generation succeeds, record:
+
+- host generated-image source path, if available;
+- copied project image path;
+- SHA256 hash when available;
+- confirmation that no review file was written.
+
+If generation cannot be completed or saved, return blocked and name the exact missing condition.
+
+## Generation Rule
+
+- Authorized task mode: prompt and generate.
+- No character image loading is required for this no-face landscape-led shot.
+- Host image generation may use standalone prompt text plus style/reference trace.
+
+## Review Rule
+
+- Review mode: none / disabled.
+- Do not write `s04-v2-review.md`.
+- Do not call Minimax MCP or any external review service.
+
+## Completion Gate Rule
+
+Before return, confirm:
+
+- version was resolved by scanning existing `s04-v*-*` files;
+- no target was overwritten;
+- brief, packet, prompt, return, and session packet/return exist;
+- image exists only if host generation and save succeeded;
+- review file is absent;
+- this worker owns only `S-04/s04-v2`.
+
+## Lifecycle Rule
+
+This worker is one-shot. After the return file is written, the worker is terminal / closed and must not be reused for repair, retry, or another slide.
+
+## Return File
+
+- Slide return: `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- Session return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`

projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md

@@ -0,0 +1,101 @@
+# S-04 Slide Prompt V2
+
+## Status
+
+- Prompt status: ready for host-native generation
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-04` / `s04-v2`
+- Overwrite allowed: no
+
+## Based On Brief
+
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- Slide function: cold morning frost and long road.
+- First-read order: cold road / frosted mud -> long unresolved road -> large empty wilderness -> optional tiny non-face traveler.
+
+## Reference Inputs
+
+- Style module: A / Landscape-Led Road.
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`.
+- Reference roles: style, composition, mood, palette, road/world-before-person grammar.
+- Common calibration reference: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`.
+- Character references: none required; if a traveler appears, keep the figure tiny, back/side, non-face, and non-identity-critical.
+- Accepted prior shot context: `S-01` validates the opening no-person road/world grammar, but this prompt should generate a new `S-04` frame rather than copy an existing shot.
+- Generation path note: standalone host-native prompt is sufficient for this no-face landscape-led shot; no visible-context character image attachment is required.
+
+## Operation Notes
+
+- Skill mode checked with `node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json`.
+- Garden/API generation is not expected by default.
+- Do not use CCPE, GPT V2 final prompts, Minimax MCP, external vision tools, or external review services.
+- Do not write `s04-v2-review.md`.
+
+## Advisor / Host-Native Record
+
+Mode check result:
+
+```json
+{
+  "mode": "B-or-C",
+  "recommendation": "host-or-advisor",
+  "garden_mode_enabled": false,
+  "has_api_key": false,
+  "base_url": "https://api.openai.com/v1",
+  "model": "gpt-image-2",
+  "env_flag_value": "(unset)"
+}
+```
+
+Skill/template files read:
+
+- `C:\Users\wangq\.agents\skills\gpt-image-2\SKILL.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\prompt-writing.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\concept-scene.md`
+- `C:\Users\wangq\.agents\skills\gpt-image-2\references\scenes-and-illustrations\minimalist-mood-scene.md`
+
+Template adaptation:
+
+- `concept-scene.md` supplies foreground / midground / background / lighting structure for a cinematic road-film still.
+- `minimalist-mood-scene.md` supplies restraint, negative space, and mood discipline.
+- Final output should be a photorealistic cinematic documentary still, not illustration, poster, fantasy concept art, or labeled board.
+
+Project source files actually read:
+
+- `AGENTS.md`
+- `docs/workflows/slide-task-orchestration.md`
+- `docs/workflows/slide-task-templates.md`
+- `projects/2026-06-23-在路上/project.md`
+- `projects/2026-06-23-在路上/execution-plan.md`
+- `projects/2026-06-23-在路上/slides/slides.md`
+- `handoff/2026-06-24-character-consistency-r4-formal-run-handoff.md`
+- `projects/2026-06-23-在路上/visual-system/visual-system.md`
+- `projects/2026-06-23-在路上/visual-system/style/style-prompt-master.md`
+- `projects/2026-06-23-在路上/visual-system/characters/character-reference-registry.md`
+
+## Final Prompt
+
+Create one 16:9 photorealistic cinematic documentary road-film still for `S-04` of the MV project `在路上`.
+
+The image must show cold morning frost and a long unresolved muddy road. The first read should be the road and the cold world, not a person. Use a landscape-led composition: a worn muddy road with wet ruts, frost along rough earth, stones, dull puddle edges, and sparse winter terrain, receding far into pale cold morning haze. The road should feel longer than the frame, continuing beyond what the viewer can see.
+
+Visual grammar: real documentary texture with restrained poetic realism. Environment first, road second, any human presence last. Low saturation palette: cold gray, earth brown, dark khaki, coal black, wet dark blue-gray. Light is weak early morning or pre-dawn silver-gray, overcast or thin haze, with only a very restrained pale edge in the sky if needed. No warm sunrise spectacle.
+
+Composition structure:
+
+- Foreground: frosted mud, wet ruts, small stones, cold puddle rims, rough road texture.
+- Midground: long road tapering through a wide empty wilderness, uneven and worn, no clean pavement.
+- Background: sparse cold land, low horizon, faint haze, no destination revealed.
+- Optional traveler: absent is acceptable. If included, make him tiny, back or side view, about `1/20` to `1/12` of the frame, unreadable face, just a small moving coordinate on the road. He must not become the subject.
+
+Camera and mood: wide horizontal road-film frame, documentary lens feeling, quiet continuation, no victory and no dramatic revelation. The image should feel like the road is still going and the person, if present, simply has to keep moving.
+
+## Negative / Avoid
+
+Avoid travel advertising, scenic tourism, postcard sunrise, broad warm horizon, clean inspirational road, centered hero backshot, portrait MV, fashion photography, polished tourism landscape, vehicle, animal, tent, camp, group, victory tone, fantasy glow, magic portal, cyberpunk, neon, sci-fi UI, code, city skyline, logos, readable text, labels, watermark, high saturation, pretty commercial landscape, or a large readable face.
+
+## Output Intent
+
+- Target path after successful host generation: `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- Review mode: disabled; no `s04-v2-review.md`.
+- If host generation or saving cannot be completed by this worker, record a blocked return rather than claiming image completion.

projects/2026-06-23-在路上/slides/s04/s04-v2-return.md

@@ -0,0 +1,76 @@
+# S-04 V2 Slide Agent Return
+
+## Status
+
+- Worker: `Agent S-04`
+- Slide: `s04`
+- Version: `s04-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: complete
+
+## Files Written
+
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s04-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`
+
+## Completeness Gate
+
+- Version resolved by scan: pass. Existing `s04` artifacts before write showed only `s04-v1-image.png`; `s04-v2` was the next append-only version.
+- No overwrite: pass. All `s04-v2-*` target files were absent before writing.
+- Expected text artifacts present: pass.
+- Expected image artifact present: pass.
+- Review file absent as instructed: pass. `projects/2026-06-23-在路上/slides/s04/s04-v2-review.md` was not written.
+- Owning worker scope: pass. This worker handled only `S-04/s04-v2`.
+- Reference loading: pass by rule. No character image reference was required because `S-04` is landscape-led and non-identity-critical.
+
+## Advisor Result
+
+- `gpt-image-2` mode: `B-or-C`
+- Recommendation: `host-or-advisor`
+- Garden mode enabled: false
+- API key present: false
+- Prompt route: Advisor / host-native
+- Skill/template files recorded in `s04-v2-prompt.md`.
+
+## Generation Result
+
+- Result: complete
+- Host generated source: `C:\Users\wangq\.codex\generated_images\019ef8ec-8032-7de0-a8b8-0d40d456fbaa\ig_059a3ef4a127f70c016a3ba1d422fc819190933818e09db431.png`
+- Project image path: `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- SHA256: `D076197E8F93821558165B4E011D7BDEB365DBA2856365E1D217B8CCE71F2E7E`
+
+## Host Generation Evidence
+
+- Garden/API expected: no
+- Host image tool used: yes
+- Visible character references loaded: not required
+- Source file preserved: yes; original host-generated PNG left in generated-images folder.
+- Project copy completed: yes
+
+## Review Result
+
+- Review mode: none / disabled by task.
+- Review file: intentionally absent.
+- External review services: not called.
+
+## Agent Lifecycle
+
+This worker is terminal / closed after this return and must not be reused for repair, retry, or another slide.
+
+## Blockers
+
+- None.
+
+## Parent Summary
+
+`S-04/s04-v2` is complete. Brief, packet, prompt, image, slide return, and mirrored session packet/return were written. The prompt was produced through the `gpt-image-2` Advisor / host-native workflow using active Module A landscape-road style guidance. No character image reference was required. Review was disabled and no review file exists.
+
+## Next Action
+
+Parent session may update the `session-05` child handoff or task log if it is aggregating S-02/S-03/S-04 status. This worker is closed.

projects/2026-06-23-在路上/slides/slides.md

@@ -28,9 +28,9 @@ Source:
 | Unit | Source Shot | Segment / Time | Function | Style Anchor | Batch | Status | Notes |
 | --- | --- | --- | --- | --- | --- | --- | --- |
 | s01 | S-01 | Prelude / 0:00-0:16 | World appears before person; cold frosted muddy road, no person. | A | anchor-test, priority | accepted | `s01-v1-image.png` accepted by user. |
-| s02 | S-02 | Prelude / 0:00-0:16 | Tiny traveler enters the world along the muddy road. | A | deferred | generated | `s02-v1-image.png` generated in broad run; awaiting batch review. |
+| s02 | S-02 | Prelude / 0:00-0:16 | Tiny traveler enters the world along the muddy road. | A | deferred | accepted / R4 complete | `s02-v1-image.png` generated in broad run; keep as historical evidence only. `s02-v2-brief.md`, `s02-v2-prompt.md`, `s02-v2-image.png`, and return exist from `session-05`; no `s02-v2-review.md`. User accepted `s02-v2-image.png`. Mechanism note: brief/packet were written by `Agent S-02` rather than prewritten by the visible child session; visible child session completed host-native generation after worker reference-attachment blocker. |
-| s03 | S-03 | Prelude / 0:00-0:16 | Old boot and backpack detail connects body to road. | character detail | deferred | generated | `s03-v1-image.png` generated in broad run; awaiting batch review. |
+| s03 | S-03 | Prelude / 0:00-0:16 | Old boot and backpack detail connects body to road. | character detail | deferred | accepted / R4 complete | `s03-v1-image.png` generated in broad run; keep as historical evidence only. `s03-v2-brief.md`, `s03-v2-prompt.md`, `s03-v2-image.png`, and return exist from `session-05`; no `s03-v2-review.md`. User accepted `s03-v2-image.png`. Mechanism note: brief/packet were written by `Agent S-03` rather than prewritten by the visible child session; visible child session completed host-native generation after worker reference-attachment blocker. |
-| s04 | S-04 | Section A / 0:16-0:32 | Cold morning frost and long road. | A | deferred | generated | `s04-v1-image.png` generated in broad run; awaiting batch review. |
+| s04 | S-04 | Section A / 0:16-0:32 | Cold morning frost and long road. | A | deferred | accepted / R4 sequencing warning | `s04-v1-image.png` generated in broad run; keep as historical evidence only. `s04-v2-brief.md`, `s04-v2-prompt.md`, `s04-v2-image.png`, and return exist from `session-05`; no `s04-v2-review.md`. User accepted `s04-v2-image.png`. Mechanism warning: `Agent S-04` was launched before the visible child session separately wrote and confirmed brief/packet; no collision, overwrite, duplicate worker, or review file was observed. |
 | s05 | S-05 | Section A / 0:16-0:32 | Backpack and coat movement lock character props. | character detail | deferred | generated / local review complete, overwrite defect recorded | `s05-v1-image.png` now points to the reference-capable rerun. Earlier broad-run `v1` candidate was overwritten during the defective workflow; do not treat this as a repeatable pattern. |
 | s06 | S-06 | Section A / 0:16-0:32 | Traveler pauses at forked muddy road. | A | deferred | accepted / mechanism pass | `s06-v1-image.png` generated in broad run; keep as candidate evidence only. `s06-v2-brief.md`, `s06-v2-prompt.md`, and `s06-v2-image.png` exist from clean `session-04` continuation with `Agent S-06` `019ef886-8797-7243-9427-d45065fcfc47`; no `s06-v2-review.md`. Prompt records `gpt-image-2` Advisor, style module A, and loaded back/walking character refs. User accepted `s06-v2-image.png`. |
 | s07 | S-07 | Section A / 0:16-0:32 | Wind/sand presses against tiny traveler. | A | deferred | generated | `s07-v1-image.png` generated in broad run; awaiting batch review. |
@@ -99,6 +99,12 @@ Accepted character-consistency workflow outputs:
 - `s33/s33-v2-image.png`
 - `s06/s06-v2-image.png`
 
+Accepted R4 formal-run outputs:
+
+- `s02/s02-v2-image.png` (accepted; brief/packet were worker-written)
+- `s03/s03-v2-image.png` (accepted; brief/packet were worker-written)
+- `s04/s04-v2-image.png` (accepted; sequencing warning recorded)
+
 Latest clean mechanism pass:
 
 - Task: `tasks/2026-06-24-character-consistency/session-04/`
@@ -107,10 +113,9 @@ Latest clean mechanism pass:
 - Review files: intentionally absent for both slides.
 - Production route: `gpt-image-2` Advisor prompt/spec plus active style and character references loaded before host image generation.
 
-Recommended R4 formal-run simulation scope:
+R4 formal-run result:
 
-- `s02` -> expected next version `s02-v2`
+- `s02-v2`, `s03-v2`, and `s04-v2` images are user accepted.
-- `s03` -> expected next version `s03-v2`
+- Review files were intentionally absent.
-- `s04` -> expected next version `s04-v2`
+- The three broad-run `v1` images remain historical evidence only, not accepted finals.
-
+- Mechanism status is not a clean pass: `S-02` and `S-03` brief/packet files were written by their slide workers rather than prewritten by the visible child session; `S-04` additionally has a sequencing warning because the worker launched before visible-child brief/packet prewrite confirmation.
-These three shots already have broad-run `v1` images only. Treat those `v1` images as historical evidence, not accepted finals, because they do not have prompt/reference/style-master traceability.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/child-session-handoff-s03-s04.md

@@ -0,0 +1,58 @@
+# Session 05 Handoff S-03 S-04
+
+## Status
+
+- Status: warning
+- Slide scope: `S-03`, then `S-04`
+- Running mode: prompt and generate
+- Review mode: none / disabled
+- Overwrite allowed: no
+
+## Completion Gate
+
+- Gate result: warning
+- File gate: pass for both slides.
+- Mechanism warning: `S-04/s04-v2` worker was launched before the visible child session had separately written and confirmed `s04-v2-brief.md` and `s04-v2-packet.md`.
+- Collision / overwrite impact: none observed. The expected S-04 brief, packet, prompt, image, and return files exist, and no review file exists.
+- Parent inspection: inspect S-04 sequencing only if deciding whether to count this as a clean mechanism pass.
+
+## Review Mode / Result
+
+- Review mode: none / disabled.
+- Review result: skipped by task rule.
+- Review files: absent for `s03-v2` and `s04-v2`.
+
+## Workers
+
+- S-03 worker id: `019ef8e2-5229-72a3-93c9-16e46a87d3bf`
+- S-03 worker lifecycle: terminal / closed after return.
+- S-04 worker id: `019ef8ec-8032-7de0-a8b8-0d40d456fbaa`
+- S-04 worker lifecycle: terminal / closed after return.
+
+## Artifacts
+
+S-03:
+
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s03-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+- SHA256: `C7FFE72E9CB4657469AD47A2701C34D35170714B77D278938B86FB7C7A6CE66A`
+
+S-04:
+
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s04-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`
+- SHA256: `D076197E8F93821558165B4E011D7BDEB365DBA2856365E1D217B8CCE71F2E7E`
+
+## Parent Summary
+
+`S-03/s03-v2` and `S-04/s04-v2` have complete expected artifact sets and no review files. `S-03` followed the intended worker boundary, with visible-session host generation after worker reference-attachment blocker. `S-04` has a sequencing warning because the worker was launched before visible-session brief/packet prewrite confirmation; no duplicate worker, collision, overwrite, review file, or extra slide work was observed.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/child-session-handoff.md

@@ -0,0 +1,77 @@
+# Session 05 Handoff
+
+## Status
+
+- Status: complete
+- Slide scope: `S-02` only
+- Target version: `s02-v2`
+- Running mode: prompt and generate
+- Review mode: none / disabled
+- Overwrite allowed: no
+
+## Completed Slides
+
+- `S-02` -> `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+
+## Completion Gate
+
+- Gate result: pass
+- Version resolved by scanning: pass. Existing `s02` artifacts before write showed only `s02-v1-image.png`; `s02-v2` was the next append-only version.
+- No overwrite: pass. New `s02-v2-*` targets were absent before writing.
+- Expected files present: pass.
+  - `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+  - `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+  - `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+  - `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+  - `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- Review file absent as instructed: pass. `projects/2026-06-23-在路上/slides/s02/s02-v2-review.md` is absent.
+- Owning worker count: pass. Exactly one internal worker owned `S-02/s02-v2`.
+- Worker lifecycle: terminal / closed after return.
+- Reference loading: pass at visible-child-session level. The internal worker blocked before text-only identity-critical generation because it could not attach local reference images; the visible child session loaded the required character references and completed the host-native generation step.
+- Generated image recorded: pass.
+  - Host generated source: `C:\Users\wangq\.codex\generated_images\019ef8cf-4ab1-7b41-877b-979dc25b0a76\ig_0c0956ca66e33513016a3b9b7510ec8191aa1013dfd66f580c.png`
+  - Project image path: `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+  - SHA256: `E35DB0100CA44314CA42EF360047BF9085C93011D10CCAB4259B724A496B9173`
+
+## Review Mode / Result
+
+- Review mode: none / disabled by task.
+- Review result: skipped by rule.
+- External review services: not called.
+
+## Agent Lifecycle
+
+- Owning worker id: `019ef8d1-188a-70c3-8a7c-4e915144a276`
+- Worker nickname: `Avicenna`
+- Worker assignment: `S-02/s02-v2` only
+- Worker final return: `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- Lifecycle status: terminal / closed after return.
+- Runtime note: the worker return records a host-reference-loading blocker. The visible child session completed only the host-native image generation and copy step using the saved prompt and loaded references; no second slide worker was launched.
+
+## Files Written
+
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s02-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/child-session-handoff.md`
+
+## Problems / Repairs Needed
+
+- None for completion gate.
+- Note for parent: worker-local host reference attachment was unavailable, so the visible child session performed the host-native generation step. This was not a duplicate-worker repair and did not create a second slide worker.
+
+## User Decisions Needed
+
+- None.
+
+## Parent Summary
+
+`S-02/s02-v2` is complete. The prompt was produced through `gpt-image-2` Advisor / host-native workflow, active A style guidance, and active walking-back character references. The image was generated by the visible child session after loading the required references, copied to the slide folder, and hashed. Review was intentionally disabled and no review file exists. Parent does not need to inspect detailed artifacts unless it wants to audit the host-reference boundary note.
+
+## Next Session Input
+
+Continue this same visible child session with the second package when instructed. Do not reuse worker `019ef8d1-188a-70c3-8a7c-4e915144a276`; launch a new one-shot worker for the next slide/version.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s02-v2-slide-agent-packet.md

@@ -0,0 +1,37 @@
+# S-02 Slide Agent Packet V2
+
+This session packet mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md
+```
+
+The authoritative packet content is kept in the slide artifact set because this worker was instructed to produce `s02-v2-packet.md` under `slides/s02/`. This task-session record exists to preserve the session-05 orchestration trail.
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-02`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-02` / `s02-v2`
+- Overwrite allowed: no
+
+## Output Targets
+
+- Brief: `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- Packet: `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+- Prompt: `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+- Image: `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+- Return: `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- Session return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+- Review: disabled; do not write `s02-v2-review.md`
+
+## Required Reference Roles
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-back-reference-crop-tight.png`
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-back-v2-from-anchor.png`
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-back-reference-crop-tight.png`
+
+If these images cannot be loaded into the visible host generation context, generation must be blocked before any text-only identity-critical output is produced.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s03-v2-slide-agent-packet.md

@@ -0,0 +1,37 @@
+# S-03 Slide Agent Packet V2
+
+This session packet mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md
+```
+
+The authoritative packet content is kept in the slide artifact set because this worker was instructed to produce `s03-v2-packet.md` under `slides/s03/`. This task-session record exists to preserve the session-05 orchestration trail.
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-03`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-03` / `s03-v2`
+- Overwrite allowed: no
+
+## Output Targets
+
+- Brief: `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- Packet: `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- Prompt: `projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md`
+- Image: `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png`
+- Return: `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- Session return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+- Review: disabled; do not write `s03-v2-review.md`
+
+## Required Reference Roles
+
+- `Image 1`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-codex-posture-test.png`
+- `Image 2`: `projects/2026-06-23-在路上/visual-system/characters/character-anchor-v2-walk-side-reference-crop-tight.png`
+- `Image 3`: `projects/2026-06-23-在路上/visual-system/characters/character-slice-walk-side-v2-from-anchor.png`
+
+If these images cannot be loaded into the visible host generation context, generation must be blocked before any text-only prop / identity-critical output is produced.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s04-v2-slide-agent-packet.md

@@ -0,0 +1,38 @@
+# S-04 Slide Agent Packet V2
+
+This session packet mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md
+```
+
+The authoritative packet content is kept in the slide artifact set because this worker was instructed to produce `s04-v2-packet.md` under `slides/s04/`. This task-session record exists to preserve the session-05 orchestration trail.
+
+## Status
+
+- Packet status: ready
+- Worker logical name: `Agent S-04`
+- Carrier type: internal one-shot slide worker
+- Running mode: `prompt and generate`
+- Review: disabled
+- Target slide/version: `S-04` / `s04-v2`
+- Overwrite allowed: no
+
+## Output Targets
+
+- Brief: `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- Packet: `projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md`
+- Prompt: `projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md`
+- Image: `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- Return: `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- Session return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`
+- Review: disabled; do not write `s04-v2-review.md`
+
+## Required Reference Roles
+
+- Style module: A / Landscape-Led Road.
+- Active style reference: `projects/2026-06-23-在路上/intake/imgs/A.风景主导型蓝图.png`.
+- Common style calibration reference: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png`.
+- Character image references: none required because `S-04` is a landscape-led no-face road shot.
+
+No visible character identity must be preserved for this slide. If a traveler appears, it must remain tiny, non-face, and non-identity-critical.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md

@@ -0,0 +1,34 @@
+# S-02 V2 Slide Agent Return
+
+This session return mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s02/s02-v2-return.md
+```
+
+## Status
+
+- Worker: `Agent S-02`
+- Slide: `s02`
+- Version: `s02-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: blocked
+- Blocker: required local character reference images could not be loaded into this worker's host image-generation context as `Image 1`, `Image 2`, and `Image 3`.
+
+## Files Written
+
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s02-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s02-v2-slide-agent-return.md`
+
+## Missing Expected File
+
+- `projects/2026-06-23-在路上/slides/s02/s02-v2-image.png`
+
+## Parent Summary
+
+Prompt artifacts are complete for `S-02/s02-v2`; host image generation is blocked until the visible parent session loads the three required character references and saves the generated PNG to the target image path. No review file was written. Worker is terminal / closed.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md

@@ -0,0 +1,34 @@
+# S-03 V2 Slide Agent Return
+
+This session return mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s03/s03-v2-return.md
+```
+
+## Status
+
+- Worker: `Agent S-03`
+- Slide: `s03`
+- Version: `s03-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: blocked
+- Blocker: required local character / prop reference images could not be loaded into this worker's host image-generation context as `Image 1`, `Image 2`, and `Image 3`.
+
+## Files Written
+
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s03-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s03-v2-slide-agent-return.md`
+
+## Missing Expected File
+
+- `projects/2026-06-23-在路上/slides/s03/s03-v2-image.png`
+
+## Parent Summary
+
+Prompt artifacts are complete for `S-03/s03-v2`; host image generation is blocked until the visible parent session loads the three required character / prop references and places the generated PNG at the target image path. No review file was written. Worker is terminal / closed.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md

@@ -0,0 +1,46 @@
+# S-04 V2 Slide Agent Return
+
+This session return mirrors:
+
+```text
+projects/2026-06-23-在路上/slides/s04/s04-v2-return.md
+```
+
+## Status
+
+- Worker: `Agent S-04`
+- Slide: `s04`
+- Version: `s04-v2`
+- Running mode: `prompt and generate`
+- Review: disabled
+- Result: complete
+
+## Files Written
+
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-brief.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-packet.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/packets/s04-v2-slide-agent-packet.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-prompt.md`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- `projects/2026-06-23-在路上/slides/s04/s04-v2-return.md`
+- `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-05/returns/s04-v2-slide-agent-return.md`
+
+## Generation Evidence
+
+- Host generated source: `C:\Users\wangq\.codex\generated_images\019ef8ec-8032-7de0-a8b8-0d40d456fbaa\ig_059a3ef4a127f70c016a3ba1d422fc819190933818e09db431.png`
+- Project image path: `projects/2026-06-23-在路上/slides/s04/s04-v2-image.png`
+- SHA256: `D076197E8F93821558165B4E011D7BDEB365DBA2856365E1D217B8CCE71F2E7E`
+
+## Completeness Gate
+
+- Version scan: pass; `s04-v2` was the next append-only version after existing `s04-v1-image.png`.
+- No overwrite: pass.
+- Expected files present: pass.
+- Review file absent as instructed: pass.
+- Character reference loading: not required for this no-face landscape-led shot.
+- Worker ownership: exactly one worker for `S-04/s04-v2`.
+- Lifecycle: terminal / closed after return.
+
+## Parent Summary
+
+`S-04/s04-v2` is complete. Prompt artifacts are complete, host-native image generation succeeded, the PNG was copied into the slide folder and hashed, and no review file was written. Worker is terminal / closed.

projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/task-log.md

@@ -56,6 +56,13 @@
 - 2026-06-24: `session-04` continuation resolved `s06-v2`, wrote `child-session-continuation-s06-plan.md`, updated `worker-launch-ledger.md`, wrote `s06-v2-brief.md` and `s06-v2-slide-agent-packet.md`, then started internal `Agent S-06` `019ef886-8797-7243-9427-d45065fcfc47`.
 - 2026-06-24: `Agent S-06` wrote `s06-v2-prompt.md`, generated/copied `s06-v2-image.png`, and wrote `session-04/returns/s06-v2-slide-agent-return.md`. No `s06-v2-review.md` was written. User confirmed second-step flow checks passed and accepted the image.
 - 2026-06-24: `session-04` closed as a mechanism pass for this scope: one clean child session, two distinct internal one-shot slide workers, append-only versions, no review files, and no duplicate worker observed.
+- 2026-06-24: Started R4 `session-05` clean child thread `019ef8cf-4ab1-7b41-877b-979dc25b0a76` for first package `S-02`; parent waited by workflow and read only compact handoff/status. User reported S-02 work time as 13m55s and child context as about 107K tokens.
+- 2026-06-24: `session-05` produced `s02-v2-image.png`; user accepted the image. Review was disabled and no `s02-v2-review.md` exists.
+- 2026-06-24: Continued the same `session-05` child thread for second package `S-03` and `S-04`; parent waited by workflow and avoided image/prompt review.
+- 2026-06-24: `session-05` produced `s03-v2-image.png` and `s04-v2-image.png`; user accepted both images. Review was disabled and no `s03-v2-review.md` or `s04-v2-review.md` exists.
+- 2026-06-24: User observed total `session-05` child context around 132K tokens after the second package; context usage is acceptable enough that brief ownership, not context pressure, became the primary mechanism concern.
+- 2026-06-24: R4 mechanism check found that `s02-v2-brief.md`, `s03-v2-brief.md`, and `s04-v2-brief.md` were generated by the slide Agents, not by the visible child session before worker launch. Evidence: each slide return lists the brief and packet under worker-written files.
+- 2026-06-24: R4 second-package handoff records a `S-04/s04-v2` sequencing warning: `Agent S-04` launched before the visible child session separately wrote and confirmed `s04-v2-brief.md` and `s04-v2-packet.md`. No collision, overwrite, duplicate worker, review file, or extra slide work was observed.
 
 ## Child Session Results
 
@@ -71,12 +78,19 @@
 - `session-04` `Agent S-33`: `019ef87a-3aa1-75a2-a97c-de8f57e234e1`; generated `s33-v2` and returned successfully.
 - `session-04` `Agent S-06`: `019ef886-8797-7243-9427-d45065fcfc47`; generated `s06-v2` and returned successfully.
 - `session-04` result: same clean child thread continued from S-33 to S-06, but S-06 used a distinct one-shot internal worker and did not reuse `Agent S-33`.
+- `session-05` child thread: `019ef8cf-4ab1-7b41-877b-979dc25b0a76`; completed R4 visual production for `s02-v2`, `s03-v2`, and `s04-v2`, but not a clean mechanism pass.
+- `session-05` `Agent S-02`: `019ef8d1-188a-70c3-8a7c-4e915144a276`; worker-generated brief/packet/prompt/return, then visible child session handled host-reference generation. User accepted image.
+- `session-05` `Agent S-03`: `019ef8e2-5229-72a3-93c9-16e46a87d3bf`; worker-generated brief/packet/prompt/return, then visible child session handled host-reference generation. User accepted image.
+- `session-05` `Agent S-04`: `019ef8ec-8032-7de0-a8b8-0d40d456fbaa`; generated `s04-v2`, user accepted image, but child handoff records a sequencing warning around worker launch before visible-child brief/packet prewrite confirmation.
 
 ## Slide Status Summary
 
 | Slide | Status | Notes |
 | --- | --- | --- |
 | `s05` | generated / local review complete, with overwrite defect recorded | `s05-v1-image.png` now points to the reference-capable rerun. Earlier broad-run `v1` candidate was overwritten during the defective workflow. Do not repeat this pattern. |
+| `s02` | accepted / R4 complete | `s02-v2` image accepted by user. No review file. Mechanism note: `Agent S-02` wrote brief/packet; visible child session completed host-native generation after worker reference-attachment blocker. |
+| `s03` | accepted / R4 complete | `s03-v2` image accepted by user. No review file. Mechanism note: `Agent S-03` wrote brief/packet; visible child session completed host-native generation after worker reference-attachment blocker. |
+| `s04` | accepted / R4 sequencing warning | `s04-v2` image accepted by user. No review file. Mechanism warning: `Agent S-04` launched before visible-child brief/packet prewrite confirmation; no collision, overwrite, duplicate worker, or extra slide work observed. |
 | `s06` | accepted / mechanism pass | `s06-v2` prompt/image/return exist from clean `session-04` continuation with `Agent S-06` `019ef886-8797-7243-9427-d45065fcfc47`, no review file. Prompt records `gpt-image-2` Advisor, style module A, and visible back/walking character refs. User accepted the image. |
 | `s21` | generated / user accepted | `session-02` remains fallback evidence only. Codex thread-level child `019ef7d2-1a7f-7a90-8c04-0098239dfe0f` resolved `s21-v3`, started separate `Agent S-21` thread `019ef7d6-3ff6-7213-b68c-fd646e39f1d2`, wrote prompt/image/return, and correctly suppressed review. User accepted the image. |
 | `s14` | accepted / mechanism defect | `s14-v2` prompt/image/return exist, no review file. Prompt records `gpt-image-2` Advisor, visible character refs, and style B. User accepted the image. Mechanism is not clean: duplicate `Agent S-14` threads `019ef803-95d7-7411-8d3c-c7f2f68a32e7` and `019ef803-328a-7e40-a0b0-0333c6b5fc6c` targeted the same v2 files; return/image source belongs to the duplicate thread folder while child handoff records the other S-14 id. |
@@ -86,5 +100,5 @@
 ## Open Decisions
 
 - Duplicate-worker prevention remains a historical defect from `session-03`, but `session-04` did not reproduce it in a two-step sequential continuation.
-- R4 formal-run simulation is the next planned check: one clean child session, recommended scope `S-02`, `S-03`, `S-04`, main-session no-audit default, child-session mandatory completion gate, and up to 3 slides in the child session.
+- R4 formal-run produced visually accepted S-02/S-03/S-04 outputs but exposed a brief-ownership mechanism issue. Decide whether visible child sessions must write and confirm `sNN-vN-brief.md` and `sNN-vN-packet.md` before launching slide Agents, especially when no GPT skeleton expansion exists.
 - Child-session launch rule is closed for this repo default: use clean user-visible Codex child/session threads for brief/orchestration work, continue the same thread only inside the approved scope, and use internal SubAgent / one-shot slide workers for each slide.