AGENT SKILL
retrospective
Use at the end of a unit of work to capture durable, non-obvious knowledge into project memory (memory/) and flag decisions worth an ADR. Turns one-off discoveries into compounding, persistent context.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-12
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- retrospective
- risk
- writes-files
- family
- documentation-memory
- profile
- core
- summary
- Persist durable, non-obvious learnings into `memory/` and flag ADR-worthy decisions
- maturity
- stable
- triggers
- [ "retrospective", "retro", "record learnings", "memory/learnings", "what did we learn", "close out the work", "lessons learned" ]
- visibility
- internal
- description
- Use at the end of a unit of work to capture durable, non-obvious knowledge into project memory (memory/) and flag decisions worth an ADR. Turns one-off discoveries into compounding, persistent context.
- auto_trigger
- true
- small_model_path
- false
02 资源文件 2
- 参考文档interview-questions.md837 B
- 参考文档prd-schema.md1.4 KB
03 SKILL.md 正文
Skill: retrospective
Purpose
Close out work by persisting what was learned so future sessions (human or agent) do
not re-derive it. Keeps memory/ a living, trustworthy asset.
Required Input
- The change or task just completed (diff, branch, or summary).
- Notable surprises, gotchas, decisions, or new conventions encountered.
Output Format
- Concise entries appended to the correct memory file:
memory/learnings.md— facts and gotchas (dated entry format).memory/patterns.md— conventions others should follow.memory/context.md— updated active focus / open threads.
- A short summary of what was recorded.
- A note on whether any decision warrants an ADR (
docs/adr/).
Execution Rules
- Record only durable, non-obvious knowledge. Skip anything already clear from the code, the README, or commit messages.
- One fact per entry; keep entries short and self-contained.
- Use absolute dates (
YYYY-MM-DD), not relative ones. - Delete or correct memory entries that turned out to be wrong.
- If a significant, hard-to-reverse decision was made, recommend writing an ADR.
- Keep all artifacts in English.
04 同仓库更多 Skills
writing-plans
Use when you have a spec or requirements for a multi-step task, before touching code
★ 0
writing-clearly-and-concisely
暂无简介
★ 0
verify_changes
暂无简介
★ 0
validate_module_structure
暂无简介
★ 0
using-git-worktrees
Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback
★ 0
ui-ux-pro-max
暂无简介
★ 0