How do I install the Git Lessons skill?

Run npx skillmds add yanacuti1121/git-lessons in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Git Lessons safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: reads secrets. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Git Lessons?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Git Lessons free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Git Lessons?

yanacuti1121 (@yanacuti1121) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Git Lessons by yanacuti1121 — Agent Skill

Name: Git Lessons
Rating: 4.2 (5 reviews)
Author: yanacuti1121

Git Lessons

Use when the user asks about past bugs, recurring mistakes, what went wrong before, or wants to avoid repeating errors. Extracts lessons from git history without needing L3 memory. Examples: 'What bugs have we fixed before?', 'Have we hit this error before?', 'What patterns keep breaking?'

yanacuti1121 10a2def 3.6 KB Updated Sep 9, 2026 2 repo stars

File contents

Git Lessons Skill

Extract actionable lessons from git history — specifically fix: commits — without needing L3 memory infrastructure. git log is L3.

When to use this skill

User asks "have we seen this error before?"
You're about to change something that has been fixed before
User wants a "lessons learned" summary for a module or area
Before touching a historically buggy area — understand what kept breaking

Workflow

Step 1 — Extract fix commits

# All fix commits (last 6 months by default)
git log --grep="^fix:" --oneline --after="6 months ago"

# Fix commits touching a specific file or dir
git log --grep="^fix:" --oneline -- path/to/file

# Fix commits for a specific area (free-text search)
git log --grep="fix.*auth\|fix.*token\|fix.*deploy" --oneline -i

Step 2 — Read the diff for significant fixes

# Full diff for a specific fix commit
git show <commit-hash>

# Just the files changed
git show <commit-hash> --name-only

Step 3 — Identify patterns

Group fix commits by:

Area: which directory/module keeps getting fixes?
Type: wrong exit code, missing check, stale assumption, format mismatch
Recurrence: same file fixed 3+ times = structural risk

Step 4 — Present as lessons

Format findings as concise rules:

LESSON: [short rule]
EVIDENCE: <commit-hash> — <commit message>
AREA: path/to/file or module name
AVOID: [specific thing not to do]

Quick queries

# Top files with the most fix commits (most fragile areas)
git log --grep="^fix:" --name-only --pretty=format: | sort | uniq -c | sort -rn | head -20

# Fix commits by author (who fixes most bugs in this area)
git log --grep="^fix:" --pretty=format:"%an %s" | sort | uniq -c | sort -rn | head -10

# Recent fixes in the last 30 days
git log --grep="^fix:" --oneline --after="30 days ago"

# Fixes that touched hooks (YAMTAM-specific)
git log --grep="^fix:" --oneline -- "*.sh" "*.js"

Constraints

Extract only, never summarize speculatively. Quote exact commit messages; do not infer intent beyond what's written.
No git blame for blame assignment. This skill is for technical pattern recognition, not attribution.
Limit depth. Default to 6 months or 50 commits — older history has diminishing relevance.
Promote critical lessons to L1. If a lesson is structural (e.g., "exit 2 required for blocking hooks"), promote it: bash .claude/scripts/add-fact.sh.

Example output

LESSON: Hook blocking requires exit 2 + hookSpecificOutput.permissionDecision
EVIDENCE: fix: correct hook output format in cost-guard and rbac-guard
AREA: core/hooks/cost-guard.sh, core/hooks/rbac-guard.sh
AVOID: Using exit 0 or {decision:"block"} — silent no-op

LESSON: grep -r regex must allow content between flag and path
EVIDENCE: fix: correct bypass test case in truth-gate test suite
AREA: core/hooks/cost-guard.sh
AVOID: Pattern 'grep.*-r[[:space:]]+\.' — misses 'grep -r pattern .'

LESSON: SCHEMA.md must be explicitly skipped in stale-facts loop
EVIDENCE: drift-check.sh fix — SCHEMA.md skip added
AREA: core/scripts/drift-check.sh
AVOID: Only skipping INDEX.md — SCHEMA.md has example dates that trigger false stale

---name: git-lessonsdescription: Use when the user asks about past bugs, recurring mistakes, what went wrong before, or wants to avoid repeating errors. Extracts lessons from git history without needing L3 memory. Examples: 'What bugs have we fixed before?', 'Have we hit this error before?', 'What patterns keep breaking?'--- # Git Lessons Skill Extract actionable lessons from git history — specifically `fix:` commits — without needing L3 memory infrastructure. `git log` is L3. ## When to use this skill - User asks "have we seen this error before?" - You're about to change something that has been fixed before - User wants a "lessons learned" summary for a module or area - Before touching a historically buggy area — understand what kept breaking ## Workflow ### Step 1 — Extract fix commits ```bash # All fix commits (last 6 months by default) git log --grep="^fix:" --oneline --after="6 months ago" # Fix commits touching a specific file or dir git log --grep="^fix:" --oneline -- path/to/file # Fix commits for a specific area (free-text search) git log --grep="fix.*auth\|fix.*token\|fix.*deploy" --oneline -i ``` ### Step 2 — Read the diff for significant fixes ```bash # Full diff for a specific fix commit git show <commit-hash> # Just the files changed git show <commit-hash> --name-only ``` ### Step 3 — Identify patterns Group fix commits by: - **Area**: which directory/module keeps getting fixes? - **Type**: wrong exit code, missing check, stale assumption, format mismatch - **Recurrence**: same file fixed 3+ times = structural risk ### Step 4 — Present as lessons Format findings as concise rules: ``` LESSON: [short rule] EVIDENCE: <commit-hash> — <commit message> AREA: path/to/file or module name AVOID: [specific thing not to do] ``` ## Quick queries ```bash # Top files with the most fix commits (most fragile areas) git log --grep="^fix:" --name-only --pretty=format: | sort | uniq -c | sort -rn | head -20 # Fix commits by author (who fixes most bugs in this area) git log --grep="^fix:" --pretty=format:"%an %s" | sort | uniq -c | sort -rn | head -10 # Recent fixes in the last 30 days git log --grep="^fix:" --oneline --after="30 days ago" # Fixes that touched hooks (YAMTAM-specific) git log --grep="^fix:" --oneline -- "*.sh" "*.js" ``` ## Constraints - **Extract only, never summarize speculatively.** Quote exact commit messages; do not infer intent beyond what's written. - **No git blame for blame assignment.** This skill is for technical pattern recognition, not attribution. - **Limit depth.** Default to 6 months or 50 commits — older history has diminishing relevance. - **Promote critical lessons to L1.** If a lesson is structural (e.g., "exit 2 required for blocking hooks"), promote it: `bash .claude/scripts/add-fact.sh`. ## Example output ``` LESSON: Hook blocking requires exit 2 + hookSpecificOutput.permissionDecision EVIDENCE: fix: correct hook output format in cost-guard and rbac-guard AREA: core/hooks/cost-guard.sh, core/hooks/rbac-guard.sh AVOID: Using exit 0 or {decision:"block"} — silent no-op LESSON: grep -r regex must allow content between flag and path EVIDENCE: fix: correct bypass test case in truth-gate test suite AREA: core/hooks/cost-guard.sh AVOID: Pattern 'grep.*-r[[:space:]]+\.' — misses 'grep -r pattern .' LESSON: SCHEMA.md must be explicitly skipped in stale-facts loop EVIDENCE: drift-check.sh fix — SCHEMA.md skip added AREA: core/scripts/drift-check.sh AVOID: Only skipping INDEX.md — SCHEMA.md has example dates that trigger false stale ```

Frequently asked questions