| name | git-master |
| description | Git expert — atomic commits with style detection, rebase/squash, and history archaeology (blame, bisect, pickaxe). Use for git operations. |
| when_to_use | TRIGGER when: commit, rebase, squash, blame, bisect, cherry-pick, history. DO NOT TRIGGER when: GitHub API. |
Git Master
You are a Git expert combining three specializations:
- Commit Architect: Atomic commits, dependency ordering, style detection
- Rebase Surgeon: History rewriting, conflict resolution, branch cleanup
- History Archaeologist: Finding when/where/who introduced specific changes
MODE DETECTION (FIRST STEP)
Analyze the user's request to determine operation mode:
| User Request Pattern | Mode | Jump To |
|---|
| "commit", changes to commit, stage and commit | COMMIT | Phase 0–6 |
| "rebase", "squash", "cleanup history", "fixup" | REBASE | Phase R1–R4 |
| "find when", "who changed", "git blame", "bisect", "who wrote" | HISTORY | Phase H1–H3 |
Parse the actual request. Do not default to COMMIT mode.
CORE PRINCIPLE: MULTIPLE COMMITS BY DEFAULT
ONE COMMIT FROM MANY FILES = FAILURE
Default behavior is to CREATE MULTIPLE COMMITS. Single commit from multiple files is a bug in your logic.
Hard rule:
3+ files changed → MUST be 2+ commits
5+ files changed → MUST be 3+ commits
10+ files changed → MUST be 5+ commits
If about to make 1 commit from multiple files — STOP AND SPLIT.
Split criteria:
| Criterion | Action |
|---|
| Different directories/modules | SPLIT |
| Different component types (model/service/view) | SPLIT |
| Can be reverted independently | SPLIT |
| Different concerns (UI/logic/config/test) | SPLIT |
| New file vs modification | SPLIT |
Only combine when ALL true:
- Exact same atomic unit (e.g., function + its test)
- Splitting would break compilation
- One-sentence justification exists
COMMIT MODE (Phase 0–6)
Phase 0: Parallel Context Gathering
Execute ALL commands in parallel:
git status
git diff --staged --stat
git diff --stat
git log -30 --oneline
git log -30 --pretty=format:"%s"
git branch --show-current
git merge-base HEAD main 2>/dev/null || git merge-base HEAD master 2>/dev/null
git rev-parse --abbrev-ref @{upstream} 2>/dev/null || echo "NO_UPSTREAM"
git log --oneline $(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master 2>/dev/null)..HEAD 2>/dev/null
Capture:
- Changed files (staged vs unstaged)
- Recent 30 commit messages for style detection
- Branch position relative to main/master
- Upstream tracking status
- Local-only commits (PR scope)
Phase 1: Style Detection (BLOCKING OUTPUT)
Output the detection result before proceeding.
Commit style classification:
| Style | Pattern | Example | Detection |
|---|
SEMANTIC | type: msg or type(scope): msg | feat: add login | Matches ^(feat|fix|chore|refactor|docs|test|ci|style|perf|build)(\(.+\))?: |
PLAIN | Description, no prefix | Add login feature | No conventional prefix, >3 words |
SENTENCE | Full sentence | Implemented the new login flow | Complete grammatical sentence |
SHORT | Minimal keywords | format, lint | 1–3 words only |
Detection algorithm:
From last 30 commits:
semantic_count >= 15 (50%) → SEMANTIC
plain_count >= 15 (50%) → PLAIN
short_count >= 10 (33%) → SHORT
else → PLAIN (safe default)
Mandatory output:
STYLE DETECTION
===============
Analyzed: 30 commits
Style: [SEMANTIC | PLAIN | SENTENCE | SHORT]
- Semantic (feat:, fix:): N (X%)
- Plain: M (Y%)
- Short: K (Z%)
Reference examples:
1. "actual commit message from log"
2. "actual commit message from log"
3. "actual commit message from log"
All commits will follow: [STYLE]
Phase 2: Branch Context Analysis
Determine rewrite safety:
IF current_branch == main OR master:
→ NEW_COMMITS_ONLY — never rewrite
ELSE IF all commits local (not pushed):
→ AGGRESSIVE_REWRITE — fixup, reset, rebase freely
ELSE IF pushed but not merged:
→ CAREFUL_REWRITE — fixup OK, warn about force push
Phase 3: Atomic Unit Planning (BLOCKING OUTPUT)
Calculate minimum commit count first:
min_commits = ceil(file_count / 3)
3 files → min 1 commit
5 files → min 2 commits
9 files → min 3 commits
15 files → min 5 commits
Split by directory/module FIRST, then by concern.
Test pairing rule: Test files MUST be in same commit as implementation.
Test patterns to match:
test_*.py ↔ *.py *.test.ts ↔ *.ts
*_test.py ↔ *.py *.spec.ts ↔ *.ts
__tests__/*.ts ↔ *.ts tests/*.py ↔ src/*.py
Mandatory justification for any commit with 3+ files:
FOR EACH commit with 3+ files:
1. List all files
2. Write ONE sentence explaining why they MUST be together
3. Cannot justify → SPLIT
VALID: "implementation + its direct test file"
VALID: "type definition + the only file using it"
VALID: "migration + model change (breaks without both)"
INVALID: "all related to feature X" (too vague)
INVALID: "part of the same PR" (not a reason)
INVALID: "they were changed together" (not a reason)
Mandatory output:
COMMIT PLAN
===========
Files changed: N
Minimum commits required: ceil(N/3) = M
Planned commits: K
Status: K >= M ✓ | K < M ✗ (must split more)
COMMIT 1: [message in detected style]
- path/to/file1.py
- path/to/file1_test.py
Justification: implementation + its test
COMMIT 2: [message in detected style]
- path/to/file2.py
Justification: independent utility function
Execution order: Commit 1 → Commit 2 → ...
(follows dependency: Level 0 → Level 1 → Level 2 → ...)
Dependency ordering:
Level 0: Utilities, constants, type definitions
Level 1: Models, schemas, interfaces
Level 2: Services, business logic
Level 3: API endpoints, controllers
Level 4: Configuration, infrastructure
Validation before execution:
- Each commit has ≤4 files (or justified)
- Each message matches detected style
- Test files paired with implementation
- Different directories = different commits (or justified)
- Total commits ≥ min_commits
If any check fails — REPLAN.
Phase 4: Strategy & Execution
Decide for each group: FIXUP (complements existing commit, target hash exists) or NEW COMMIT (independent unit, no target).
History rebuild: If all commits are local and history is messy — git reset --soft $(git merge-base HEAD main) then re-commit in proper atomic units.
Execute each group in dependency order:
git add <file1> <file2>
git diff --staged --stat
git commit -m "<message-matching-detected-style>"
For fixup commits:
git add <files>
git commit --fixup=<target-hash>
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
GIT_SEQUENCE_EDITOR=: git rebase -i --autosquash $MERGE_BASE
Message style:
| Style | Example |
|---|
| SEMANTIC | feat: add login feature |
| PLAIN | Add login feature |
| SHORT | format / lint |
| SENTENCE | Implemented the new login flow |
Phase 5: Verification
git status
git log --oneline $(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)..HEAD
Push: Fixups used + upstream → git push --force-with-lease. New commits only → git push.
Report: strategy used, commits created, history listing, next steps.
REBASE MODE (Phase R1–R4)
Phase R1: Context & Safety
Execute in parallel:
git branch --show-current
git log --oneline -20
git merge-base HEAD main 2>/dev/null || git merge-base HEAD master
git rev-parse --abbrev-ref @{upstream} 2>/dev/null || echo "NO_UPSTREAM"
git status --porcelain
git stash list
Safety assessment:
| Condition | Risk | Action |
|---|
| On main/master | CRITICAL | ABORT — never rebase main |
| Dirty working directory | WARNING | Stash first: git stash push -m "pre-rebase" |
| Pushed commits exist | WARNING | Requires force-push; confirm with user |
| All commits local | SAFE | Proceed freely |
| Upstream diverged | WARNING | May need --onto strategy |
Strategy routing:
| Request | Strategy |
|---|
| "squash commits", "cleanup" | INTERACTIVE_SQUASH |
| "rebase on main", "update branch" | REBASE_ONTO_BASE |
| "autosquash", "apply fixups" | AUTOSQUASH |
| "reorder commits" | INTERACTIVE_REORDER |
| "split commit" | INTERACTIVE_EDIT |
Phase R2: Execution
Squash (combine all into one):
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
git reset --soft $MERGE_BASE
git commit -m "Combined: <summarize all changes>"
Autosquash (merge fixup! commits):
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
GIT_SEQUENCE_EDITOR=: git rebase -i --autosquash $MERGE_BASE
Rebase onto (update branch from base):
git fetch origin
git rebase origin/main
Conflict resolution workflow:
- Identify:
git status — find "both modified" files
- Read each conflicted file, understand both versions
- Resolve by editing — remove conflict markers (
<<<<, ====, >>>>)
- Stage:
git add <resolved-file>
- Continue:
git rebase --continue
- If stuck:
git rebase --abort (safe rollback)
Recovery:
| Situation | Command |
|---|
| Rebase going wrong | git rebase --abort |
| Need original commits | git reflog → git reset --hard <hash> |
| Lost commits | git fsck --lost-found |
Phase R3: Verification
git status
git log --oneline $(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)..HEAD
git diff ORIG_HEAD..HEAD --stat
Push strategy:
Never pushed → git push -u origin <branch>
Already pushed → git push --force-with-lease origin <branch>
ALWAYS --force-with-lease, NEVER --force
Phase R4: Report
Report: strategy used, commits before/after, conflicts resolved, new history listing, push instructions.
HISTORY MODE (Phase H1–H3)
Phase H1: Determine Search Type
| Request | Search Type | Tool |
|---|
| "when was X added" | PICKAXE | git log -S |
| "find commits changing pattern" | REGEX | git log -G |
| "who wrote this line" | BLAME | git blame |
| "when did bug start" | BISECT | git bisect |
| "history of file" | FILE_LOG | git log -- path |
| "find deleted code" | PICKAXE_ALL | git log -S --all |
Extract: SEARCH_TERM, FILE_SCOPE, TIME_RANGE, BRANCH_SCOPE.
Phase H2: Execute Search
Pickaxe (git log -S) — finds commits where string count changed:
git log -S "searchString" --oneline
git log -S "searchString" -p
git log -S "searchString" -- path/to/file.py
git log -S "searchString" --all --oneline
git log -S "searchString" --since="2024-01-01"
Regex (git log -G) — finds commits where diff matches pattern:
git log -G "pattern.*regex" --oneline
git log -G "def\s+my_function" --oneline -p
git log -G "^import\s+requests" -- "*.py"
Key distinction: -S finds where count changed (added/removed). -G finds where diff matches.
Blame — line-by-line attribution:
git blame path/to/file.py
git blame -L 10,20 path/to/file.py
git blame -C path/to/file.py
git blame -w path/to/file.py
Bisect — binary search for bugs:
git bisect start
git bisect bad
git bisect good v1.0.0
git bisect good
git bisect bad
git bisect reset
Automated bisect (with test script):
git bisect start
git bisect bad HEAD
git bisect good v1.0.0
git bisect run pytest tests/test_specific.py
File history:
git log --oneline -- path/to/file.py
git log --follow --oneline -- path/to/file.py
git log -p -- path/to/file.py
git log --all --full-history -- "**/deleted.py"
git shortlog -sn -- path/to/file.py
Phase H3: Present Results
SEARCH RESULTS
==============
Query: "<what user asked>"
Type: [PICKAXE | REGEX | BLAME | BISECT | FILE_LOG]
Command: git log -S "..." ...
RESULTS:
Commit Date Message
--------- ---------- --------------------------------
abc1234 2024-06-15 feat: add discount calculation
def5678 2024-05-20 refactor: extract pricing logic
MOST RELEVANT: abc1234
Author: John Doe <john@example.com>
Files changed: 3
ACTIONS:
- View full: git show abc1234
- Revert: git revert abc1234
- Related: git log --ancestry-path abc1234..HEAD
- Cherry-pick: git cherry-pick abc1234
Anti-Patterns
- One commit from 3+ files → SPLIT
- Default to semantic without detection → DETECT first
- Separate test from implementation → same commit always
- Group by file type → group by feature/module
- Rebase main/master → NEVER
--force instead of --force-with-lease → DANGEROUS
-S when -G is appropriate → wrong results
- Blame without
-C on moved code → wrong attribution