AGENT SKILL

task-summary

Focused post-task summary covering architectural decisions, decision forks, pitfalls, and plan deviations. Use at `document → review`, before handing the card to the reviewer, on a task that failed, or when the supervisor asks what happened with a task.

License 许可
MIT
所属仓库
muratovv/ai-hats
最近更新
2026-11-05
Stars
★ 3
Forks
1

01 前言 Front Matter

name
task-summary
ai_hats
{ "type": "skill", "schema_version": 1 }
license
MIT
description
Focused post-task summary covering architectural decisions, decision forks, pitfalls, and plan deviations. Use at `document → review`, before handing the card to the reviewer, on a task that failed, or when the supervisor asks what happened with a task.

03 SKILL.md 正文

Task Summary

Produce a focused summary of a completed task, capturing only decision-critical knowledge.

When to Use

The backward-looking factual record of a finished task — what was decided and why. It is due at document → review: summary.md is what the reviewer opens, so it is attached before the card enters review, not after it leaves (see hatrack's per-edge table). Two siblings to keep distinct:

  • self-retrospective is the improvement analysis (5-whys, systemic fixes) of how the work went — run that when there were failures or backtracks; this skill just records the outcome.
  • context-handoff is the forward brief for whoever continues the work — task-summary looks back, handoff looks ahead.

Procedure

  1. Identify the task: Locate the task card in <ai_hats_dir>/tracker/backlog/tasks/<ID>/.

  2. Extract decision-critical information:

    • Architectural decisions — what structural choices were made and WHY
    • Decision forks — where were there multiple viable options? Document: what options existed, which was chosen, and the key reason
    • Pitfalls discovered — non-obvious traps that cost time or could bite again
    • Deviations from plan — where did execution diverge from the original plan and why
  3. Draft the summary at a scratch path outside the backlog — e.g. /tmp/<ID>-summary.md. Do not write it under tasks/<ID>/ yourself; step 4 puts it there:

# Summary: <task title>

Completed: YYYY-MM-DD
Result: <done | failed>

## What Was Done

<2-5 sentences. Outcome, not play-by-play.>

## Key Decisions

| Decision | Why This Way | Alternative Rejected |
| -------- | ------------ | -------------------- |
| ...      | ...          | ...                  |

## Pitfalls

- <Non-obvious trap> — <impact or how to avoid>

## Deviations from Plan

- <What changed> — <why>
  1. Attach it to the card:

    rack transition <ID> --attach /tmp/<ID>-summary.md:summary.md
    

    The file is copied to tasks/<ID>/summary.md — a document, discovered live by rack context (name + digest computed on the fly, fs-as-truth). --attach is the sanctioned way in: the backlog gate refuses a hand edit under tasks/<ID>/**, carving out plan.md alone. Re-attaching the same name overwrites (the op notes it); --freeze summary.md pins its digest.

What NOT to Include

  • Step-by-step log of actions (that's work_log in the task card)
  • Routine operations (git commits, file reads, linting)
  • Information already in the code or commit messages
  • Full error traces (reference file paths instead)

Completion

  • Summary drafted outside the backlog tree, then attached via rack transition --attach
  • Contains at least: What Was Done, Key Decisions
  • rack context <ID> lists summary.md under its documents

Anti-Patterns

  • Summarizing everything — the point is filtering, not transcription
  • Missing the WHY — a list of facts without reasoning helps nobody
  • Writing the summary without reading the task's work_log first — you'll miss context
  • Writing the summary under tasks/<ID>/ by hand — summary.md there, or a summary_file: key in task.yaml. Both break the backlog gate, and no CLI writes either. rack transition <ID> --attach is the door

04 同仓库更多 Skills

muratovv
muratovv/ai-hats

tool-evaluation-protocol

Time-bounded protocol for evaluating new tools, libraries, or methodologies before adoption. Use when evaluating a candidate dependency for the project, researching a methodology (e.g. ML/LLM frameworks, testing approaches), or following up on a hype-driven recommendation.

★ 3
muratovv
muratovv/ai-hats

systematic-debugging

4-phase bug-fix protocol (evidence, pattern, hypothesis, verify). Use when investigating any error, failure, or unexpected behavior, handling bug reports from users or monitoring, or diagnosing test failures with non-obvious causes.

★ 3
muratovv
muratovv/ai-hats

skill-template

Canonical template and validation guide for ai-hats skills. Use when creating a new skill, reviewing an existing one for structural compliance, or deciding which pattern fits a new behavior.

Skill与智能体开发
★ 3
muratovv
muratovv/ai-hats

skill-optimization

Audit and refactor library components to eliminate redundancy and staleness. Use during periodic library maintenance, after adding multiple new skills or rules, when a role's token budget is too high, or after retrospectives identify component issues.

Skill与智能体开发
★ 3
muratovv
muratovv/ai-hats

silent-fallback-gate

PreToolUse hook — refuses an Edit or Write that adds an inert broad except. Use when a Python edit was blocked for a silent fallback.

★ 3
muratovv
muratovv/ai-hats

retro-to-framework

Convert an observation about agent behavior — from a retrospective or from a single incident — into a framework-level change (hook, rule, skill, skill update). Use after a retrospective identifies problems that are not project-specific, after an incident where an agent misapplied a component it had read, when CLAUDE.md band-aids accumulate (more than 3 per-project rules that could be generic), or when the same problem recurs across multiple projects.

Skill与智能体开发
★ 3