AGENT SKILL

knowledge-capture

Extract durable learnings and questions after meaningful work.

License 许可
MIT
最近更新
2026-08-08
Stars
★ 1
Forks
0

01 前言 Front Matter

name
knowledge-capture
description
Extract durable learnings and questions after meaningful work.

03 SKILL.md 正文

Knowledge Capture

Purpose

Separate active task state from durable knowledge proposals.

Extract

  • decisions made
  • assumptions introduced
  • assumptions disproved
  • new domain relationships
  • undocumented behaviours
  • integration quirks
  • failure modes
  • reusable patterns
  • contract changes
  • operational risks
  • unanswered questions
  • contradictions with existing guidance

Destination Decision

active resume state       -> HANDOFF.toon
unanswered material issue -> knowledge/questions
evidence-backed discovery -> knowledge/learnings or inbox
durable decision          -> decision proposal
reusable practice         -> pattern proposal
failure mode              -> risk proposal
trivial implementation    -> no durable record

Rules

  • Do not write directly to canonical knowledge unless reviewed evidence already exists or the user explicitly directs promotion.
  • Prefer .agents/knowledge/inbox/ for unreviewed proposals.
  • Link proposals to existing IDs when possible.
  • Avoid recording trivial implementation details.
  • Capture contradictions with existing guidance.
  • Update HANDOFF.toon.knowledge on every meaningful task: record consulted IDs/paths, proposals created, or a concrete no_record reason.

Output

handoff_updates: []
knowledge_proposals:
  - path: .agents/knowledge/inbox/INBOX-001-example.md
    reason: possible reusable pattern
questions_created: []
no_record:
  - reason: local-only mechanical change

04 同仓库更多 Skills