98 lines
2.7 KiB
Markdown
98 lines
2.7 KiB
Markdown
# Prompt To Send: topic-discovery-router
|
|
|
|
## 0. Status
|
|
|
|
```text
|
|
status: draft_not_ready_for_external_execution
|
|
reason: external execution requires either pasted canonical Agent spec plus source context pack, or a participant that can read the local file paths below
|
|
```
|
|
|
|
## 1. Canonical Agent Prompt Or Spec
|
|
|
|
Use the canonical Agent spec:
|
|
|
|
```text
|
|
C:\Users\wangq\Documents\Codex\ccpe-system\agents\agent-specs\topic-discovery-router.agent.md
|
|
```
|
|
|
|
If the participant cannot read local files, paste the full canonical Agent spec here before execution.
|
|
|
|
## 2. Runtime Wrapper
|
|
|
|
You are being invoked for one stage of a larger Knowledge Vault workflow. Preserve your canonical `topic-discovery-router` role.
|
|
|
|
Do not act as:
|
|
|
|
```text
|
|
- evidence-routing-and-topic-doc-builder
|
|
- lossless-coverage-auditor
|
|
- writing-workbench
|
|
- model-card promoter
|
|
```
|
|
|
|
## 3. Current Task
|
|
|
|
Create a candidate topic-map draft from the segmented source block inventory and enough source content to classify blocks.
|
|
|
|
## 4. Project Context
|
|
|
|
```text
|
|
project_id: 2026-06-05-失望性情感隔离
|
|
workflow: viewpoint-discussion-distillation
|
|
source_document_path: diary/2026/2026-06/2026-06-05.md
|
|
source_map_path: discussions/viewpoint-development/2026-06-05-失望性情感隔离/00-source-map.md
|
|
source_block_inventory: discussions/viewpoint-development/2026-06-05-失望性情感隔离/invocations/conversation-segmentation/source-block-inventory.csv
|
|
round_count: 7
|
|
source_block_count: 40
|
|
```
|
|
|
|
## 5. Input Files Or Embedded Inputs
|
|
|
|
If running locally, read:
|
|
|
|
```text
|
|
00-source-map.md
|
|
invocations/conversation-segmentation/source-block-inventory.csv
|
|
diary/2026/2026-06/2026-06-05.md
|
|
```
|
|
|
|
If running externally, the operator must provide a source context pack containing source excerpts by block id. Do not classify from locator text alone.
|
|
|
|
## 6. Context Pack
|
|
|
|
Pending user confirmation:
|
|
|
|
```text
|
|
option_a: one external/local participant reads all 40 source blocks
|
|
option_b: seven worker participants process R01-R07 and return partial topic candidates
|
|
option_c: main session prepares source excerpt packs by block batch, then user runs external topic-discovery-router
|
|
```
|
|
|
|
## 7. Output Requirements
|
|
|
|
Return Markdown with:
|
|
|
|
```text
|
|
1. Topic Summary
|
|
2. Topic Scope And Anti-Scope
|
|
3. Source Block Routing
|
|
4. Downstream Task Candidates
|
|
5. Merge / Split Proposals
|
|
6. Low-Confidence Items
|
|
7. Human Decisions Needed
|
|
```
|
|
|
|
Every topic must include source block IDs and confidence.
|
|
|
|
## 8. Session Continuity Instructions
|
|
|
|
Do not rely on chat memory. Use file-backed source block ids and return output to:
|
|
|
|
```text
|
|
invocations/topic-discovery-router/returned-output.md
|
|
```
|
|
|
|
## 9. Return Instructions
|
|
|
|
Return only the formal topic-discovery-router output and any explicit uncertainty. Do not generate topic documents.
|