Agent Skills
› microsoft/Generative-AI-for-beginners-dotnet
› history-hygiene
history-hygiene
GitHub规范历史记录的维护，要求仅记录最终确认的结果而非中间状态或已撤销的决定，确保未来智能体读取时获得唯一且准确的事实依据，避免决策污染。
Trigger Scenarios
需要更新项目历史记录文件时
决策发生变更或反转时
Install
npx skills add microsoft/Generative-AI-for-beginners-dotnet --skill history-hygiene -g -y
SKILL.md
Frontmatter
{
"name": "history-hygiene",
"domain": "documentation, team-collaboration",
"source": "earned (Kobayashi v0.6.0 incident, team intervention)",
"confidence": "high",
"description": "Record final outcomes to history.md, not intermediate requests or reversed decisions"
}
Context
History files (.md files tracking decisions, spawns, outcomes) are read cold by future agents. Stale or incorrect entries poison decision-making downstream. The Kobayashi incident proved this: history said "Brady decided v0.6.0" when Brady had reversed that to v0.8.17. Future spawns read the wrong truth and repeated the mistake.
Patterns
- Record the final outcome, not the initial request.
- Wait for confirmation before writing to history — don't log intermediate states.
- If a decision reverses, update the entry immediately — don't leave stale data.
- One read = one truth. A future agent should never need to cross-reference other files to understand what actually happened.
Examples
✓ Correct:
- "Migration target: v0.8.17 (initially discussed as v0.6.0, corrected by Brady)"
- "Reverted to Node 18 per Brady's explicit request on 2024-01-15"
✗ Incorrect:
- "Brady directed v0.6.0" (when later reversed)
- Recording what was requested instead of what actually happened
- Logging entries before outcome is confirmed
Anti-Patterns
- Writing intermediate or "for now" states to disk
- Attributing decisions without confirming final direction
- Treating history like a draft — history is the source of truth
- Assuming readers will cross-reference or verify; they won't
Version History
- bd4e082 Current 2026-08-20 14:06
