# S-16 Slide Agent Packet V3

## Status

- Status: ready.
- Packet owner: visible child session, `session-09`.
- Worker type: internal one-shot worker.
- Worker profile: `video-slide-production`.
- Running mode: `prompt and generate`.
- Review: none / intentionally suppressed.
- Overwrite allowed: no.

## Target Slide

- Slide: `S-16`.
- Version: `s16-v3`.
- Brief: `projects/2026-06-23-在路上/slides/s16/s16-v3-brief.md`.

## Allowed Actions

- Read the brief and this packet.
- Read active project and visual-system source files named by the brief.
- Run or record the fresh `gpt-image-2` mode check.
- Use `gpt-image-2` as Advisor / host-native prompt workflow.
- Generate one 16:9 image through host-native image generation.
- Write only:
  - `projects/2026-06-23-在路上/slides/s16/s16-v3-prompt.md`
  - `projects/2026-06-23-在路上/slides/s16/s16-v3-image.png`
  - `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-09/returns/s16-v3-slide-agent-return.md`

## Forbidden Actions

- Do not create, edit, or rewrite any brief file.
- Do not create, edit, or rewrite any packet file.
- Do not write `projects/2026-06-23-在路上/slides/s16/s16-v3-review.md`.
- Do not process any slide other than `S-16`.
- Do not call Minimax MCP, CCPE, GPT V2, or external review/vision services.
- Do not overwrite any existing file.

## 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`
- `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/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/slides/s16/s16-v2-prompt.md`
- `projects/2026-06-23-在路上/slides/s16/s16-v3-brief.md`

## Output Files

- Prompt: `projects/2026-06-23-在路上/slides/s16/s16-v3-prompt.md`
- Image: `projects/2026-06-23-在路上/slides/s16/s16-v3-image.png`
- Return: `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-09/returns/s16-v3-slide-agent-return.md`
- Review: intentionally absent; do not create `projects/2026-06-23-在路上/slides/s16/s16-v3-review.md`

## Artifact Version / Write Policy

- This is append-only `s16-v3`.
- Stop with a blocked return if prompt, image, return, or review target already exists.
- If image generation or file copying partially fails after writing any output, preserve the incomplete `s16-v3` evidence and do not complete it in place.

## Reference Roles

These images are supplied in the worker context:

- Image 1: `projects/2026-06-23-在路上/visual-system/style/style-anchor-board-v1.png` as accepted style calibration; use Module A as the positive style direction.
- Image 2: `projects/2026-06-23-在路上/slides/s15/s15-v2-image.png` as negative contrast; avoid repeating its composition/person scale relationship.
- Image 3: `projects/2026-06-23-在路上/slides/s16/s16-v2-image.png` as failed S-16 negative contrast; avoid repeating its insufficiently extreme road/person relation.

Character identity images are not required for this version because the traveler must be extremely small. Do not enlarge the traveler to preserve face/body detail.

## Visual-System Requirements

- Use style module `A / Landscape-Led Road`.
- Preserve the global kernel: real documentary texture + restrained poetic realism + landscape-led road-film composition.
- Required repair: horizontal extreme wide shot; road cuts from one side toward far distance; traveler extremely small.

## Gpt-Image-2 Advisor Rule

- Mode check command: `node C:\Users\wangq\.agents\skills\gpt-image-2\scripts\check-mode.js --json`.
- Expected mode: `B-or-C`; effective path is Mode B host-native if `image_gen` is available.
- Skill files already read by visible child session:
  - `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`
- The prompt file must record the Skill mode, template/reference paths, and project source files used.

## Host Generation Evidence Rule

- Garden/API expected: no.
- Use host-native `image_gen`.
- Actual Image 1 / Image 2 / Image 3 references are visible in worker context.
- Save/copy the generated image to the target project path and record the source path, project path, dimensions/size if available, and SHA256.

## Generation Rule

- Generate a 16:9 horizontal image.
- The final image must prioritize road geometry and scale over identity detail.
- If the traveler is more than a tiny coordinate, the run should be treated as visually suspect in the return summary.

## Review Rule

- Review mode: none.
- Do not write a review file.
- Return must state review was intentionally suppressed and the review file is absent.

## Completion Gate Rule

Before returning, confirm:

- prompt/image/return exist;
- no target file was overwritten;
- no review file exists;
- brief and packet were read-only inputs and were not written by the worker;
- Image 1 / Image 2 / Image 3 were available in the worker context, or the run blocked before generation;
- worker handled only `S-16/s16-v3`.

## Lifecycle Rule

This worker is terminal after writing the return. It must not be continued or reused.

## Return File

Write `projects/2026-06-23-在路上/tasks/2026-06-24-character-consistency/session-09/returns/s16-v3-slide-agent-return.md`.