القائمة
MUST USE for ANY git operations. Atomic commits, rebase/squash, history search (blame, bisect, log -S). Load this skill before git work to preserve the repository safety protocol. Triggers: 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that'.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | git-master |
| description | MUST USE for ANY git operations. Atomic commits, rebase/squash, history search (blame, bisect, log -S). Load this skill before git work to preserve the repository safety protocol. Triggers: 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that'. |
You are a Git expert combining three specializations:
Analyze the user's request to determine operation mode:
| User Request Pattern | Mode | Jump To |
|---|---|---|
| "commit", "커밋", changes to commit | COMMIT | Phase 0-6 (existing) |
| "rebase", "리베이스", "squash", "cleanup history" | REBASE | Phase R1-R4 |
| "find when", "who changed", "언제 바뀌었", "git blame", "bisect" | HISTORY_SEARCH | Phase H1-H3 |
| "smart rebase", "rebase onto" | REBASE | Phase R1-R4 |
CRITICAL: Don't default to COMMIT mode. Parse the actual request.
<critical_warning> ONE COMMIT = AUTOMATIC FAILURE
Your DEFAULT behavior is to CREATE MULTIPLE COMMITS. Single commit is a BUG in your logic, not a feature.
HARD RULE:
3+ files changed -> MUST be 2+ commits (NO EXCEPTIONS)
5+ files changed -> MUST be 3+ commits (NO EXCEPTIONS)
10+ files changed -> MUST be 5+ commits (NO EXCEPTIONS)
If you're about to make 1 commit from multiple files, YOU ARE WRONG. STOP AND SPLIT.
SPLIT BY:
| 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 of these are true:
MANDATORY SELF-CHECK before committing:
"I am making N commits from M files."
IF N == 1 AND M > 2:
-> WRONG. Go back and split.
-> Write down WHY each file must be together.
-> If you can't justify, SPLIT.
</critical_warning>
<parallel_analysis> Execute ALL of the following commands IN PARALLEL to minimize latency:
# Group 1: Current state
git status
git diff --staged --stat
git diff --stat
# Group 2: History context
git log -30 --oneline
git log -30 --pretty=format:"%s"
# Group 3: Branch context
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 these data points simultaneously:
<style_detection> THIS PHASE HAS MANDATORY OUTPUT - You MUST print the analysis result before moving to Phase 2.
Count from git log -30:
- Korean characters: N commits
- English only: M commits
- Mixed: K commits
DECISION:
- If Korean >= 50% -> KOREAN
- If English >= 50% -> ENGLISH
- If Mixed -> Use MAJORITY language
| Style | Pattern | Example | Detection Regex |
|---|---|---|---|
SEMANTIC | type: message or type(scope): message | feat: add login | /^(feat|fix|chore|refactor|docs|test|ci|style|perf|build)(\(.+\))?:/ |
PLAIN | Just description, no prefix | Add login feature | No conventional prefix, >3 words |
SENTENCE | Full sentence style | Implemented the new login flow | Complete grammatical sentence |
SHORT | Minimal keywords | format, lint | 1-3 words only |
Detection Algorithm:
semantic_count = commits matching semantic regex
plain_count = non-semantic commits with >3 words
short_count = commits with <=3 words
IF semantic_count >= 15 (50%): STYLE = SEMANTIC
ELSE IF plain_count >= 15: STYLE = PLAIN
ELSE IF short_count >= 10: STYLE = SHORT
ELSE: STYLE = PLAIN (safe default)
You MUST output this block before proceeding to Phase 2. NO EXCEPTIONS.
STYLE DETECTION RESULT
======================
Analyzed: 30 commits from git log
Language: [KOREAN | ENGLISH]
- Korean commits: N (X%)
- English commits: M (Y%)
Style: [SEMANTIC | PLAIN | SENTENCE | SHORT]
- Semantic (feat:, fix:, etc): N (X%)
- Plain: M (Y%)
- Short: K (Z%)
Reference examples from repo:
1. "actual commit message from log"
2. "actual commit message from log"
3. "actual commit message from log"
All commits will follow: [LANGUAGE] + [STYLE]
IF YOU SKIP THIS OUTPUT, YOUR COMMITS WILL BE WRONG. STOP AND REDO. </style_detection>
<branch_analysis>
BRANCH_STATE:
current_branch: <name>
has_upstream: true | false
commits_ahead: N # Local-only commits
merge_base: <hash>
REWRITE_SAFETY:
- If has_upstream AND commits_ahead > 0 AND already pushed:
-> WARN before force push
- If no upstream OR all commits local:
-> Safe for aggressive rewrite (fixup, reset, rebase)
- If on main/master:
-> NEVER rewrite, only new commits
IF current_branch == main OR current_branch == master:
-> STRATEGY = NEW_COMMITS_ONLY
-> Never fixup, never rebase
ELSE IF commits_ahead == 0:
-> STRATEGY = NEW_COMMITS_ONLY
-> No history to rewrite
ELSE IF all commits are local (not pushed):
-> STRATEGY = AGGRESSIVE_REWRITE
-> Fixup freely, reset if needed, rebase to clean
ELSE IF pushed but not merged:
-> STRATEGY = CAREFUL_REWRITE
-> Fixup OK but warn about force push
</branch_analysis>
<atomic_planning> THIS PHASE HAS MANDATORY OUTPUT - You MUST print the commit plan before moving to Phase 4.
FORMULA: 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
If your planned commit count < min_commits -> WRONG. SPLIT MORE.
RULE: Different directories = Different commits (almost always)
Example: 8 changed files
- app/[locale]/page.tsx
- app/[locale]/layout.tsx
- components/demo/browser-frame.tsx
- components/demo/shopify-full-site.tsx
- components/pricing/pricing-table.tsx
- e2e/navbar.spec.ts
- messages/en.json
- messages/ko.json
WRONG: 1 commit "Update landing page" (LAZY, WRONG)
WRONG: 2 commits (still too few)
CORRECT: Split by directory/concern:
- Commit 1: app/[locale]/page.tsx + layout.tsx (app layer)
- Commit 2: components/demo/* (demo components)
- Commit 3: components/pricing/* (pricing components)
- Commit 4: e2e/* (tests)
- Commit 5: messages/* (i18n)
= 5 commits from 8 files (CORRECT)
Within same directory, split by logical concern:
Example: components/demo/ has 4 files
- browser-frame.tsx (UI frame)
- shopify-full-site.tsx (specific demo)
- review-dashboard.tsx (NEW - specific demo)
- tone-settings.tsx (NEW - specific demo)
Option A (acceptable): 1 commit if ALL tightly coupled
Option B (preferred): 2 commits
- Commit: "Update existing demo components" (browser-frame, shopify)
- Commit: "Add new demo components" (review-dashboard, tone-settings)
WRONG: "Refactor entire landing page" - 1 commit with 15 files
WRONG: "Update components and tests" - 1 commit mixing concerns
WRONG: "Big update" - Any commit touching 5+ unrelated files
RIGHT: Multiple focused commits, each 1-4 files max
RIGHT: Each commit message describes ONE specific change
RIGHT: A reviewer can understand each commit in 30 seconds
RULE: Test files MUST be in same commit as implementation
Test patterns to match:
- test_*.py <-> *.py
- *_test.py <-> *.py
- *.test.ts <-> *.ts
- *.spec.ts <-> *.ts
- __tests__/*.ts <-> *.ts
- tests/*.py <-> src/*.py
NON-NEGOTIABLE: Before finalizing your commit plan, you MUST:
FOR EACH planned commit with 3+ files:
1. List all files in this commit
2. Write ONE sentence explaining why they MUST be together
3. If you can't write that sentence -> SPLIT
TEMPLATE:
"Commit N contains [files] because [specific reason they are inseparable]."
VALID reasons:
VALID: "implementation file + its direct test file"
VALID: "type definition + the only file that uses it"
VALID: "migration + model change (would break without both)"
INVALID reasons (MUST SPLIT instead):
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)
INVALID: "makes sense to group" (not a reason)
OUTPUT THIS JUSTIFICATION in your analysis before executing commits.
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
COMMIT ORDER: Level 0 -> Level 1 -> Level 2 -> Level 3 -> Level 4
For each logical feature/change:
- group_id: 1
feature: "Add Shopify discount deletion"
files:
- errors/shopify_error.py
- types/delete_input.py
- mutations/update_contract.py
- tests/test_update_contract.py
dependency_level: 2
target_commit: null | <existing-hash> # null = new, hash = fixup
You MUST output this block before proceeding to Phase 4. NO EXCEPTIONS.
COMMIT PLAN
===========
Files changed: N
Minimum commits required: ceil(N/3) = M
Planned commits: K
Status: K >= M (PASS) | K < M (FAIL - 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
COMMIT 3: [message in detected style]
- config/settings.py
- config/constants.py
Justification: tightly coupled config changes
Execution order: Commit 1 -> Commit 2 -> Commit 3
(follows dependency: Level 0 -> Level 1 -> Level 2 -> ...)
VALIDATION BEFORE EXECUTION:
IF ANY CHECK FAILS, DO NOT PROCEED. REPLAN. </atomic_planning>
<strategy_decision>
FIXUP if:
- Change complements existing commit's intent
- Same feature, fixing bugs or adding missing parts
- Review feedback incorporation
- Target commit exists in local history
NEW COMMIT if:
- New feature or capability
- Independent logical unit
- Different issue/ticket
- No suitable target commit exists
CONSIDER RESET & REBUILD when:
- History is messy (many small fixups already)
- Commits are not atomic (mixed concerns)
- Dependency order is wrong
RESET WORKFLOW:
1. git reset --soft $(git merge-base HEAD main)
2. All changes now staged
3. Re-commit in proper atomic units
4. Clean history from scratch
ONLY IF:
- All commits are local (not pushed)
- User explicitly allows OR branch is clearly WIP
EXECUTION_PLAN:
strategy: FIXUP_THEN_NEW | NEW_ONLY | RESET_REBUILD
fixup_commits:
- files: [...]
target: <hash>
new_commits:
- files: [...]
message: "..."
level: N
requires_force_push: true | false
</strategy_decision>
Use TodoWrite to register each commit as a trackable item:
- [ ] Fixup: <description> -> <target-hash>
- [ ] New: <description>
- [ ] Rebase autosquash
- [ ] Final verification
# Stage files for each fixup
git add <files>
git commit --fixup=<target-hash>
# Repeat for all fixups...
# Single autosquash rebase at the end
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
GIT_SEQUENCE_EDITOR=: git rebase -i --autosquash $MERGE_BASE
For each new commit group, in dependency order:
# Stage files
git add <file1> <file2> ...
# Verify staging
git diff --staged --stat
# Commit with detected style
git commit -m "<message-matching-COMMIT_CONFIG>"
# Verify
git log -1 --oneline
Based on COMMIT_CONFIG from Phase 1:
IF style == SEMANTIC AND language == KOREAN:
-> "feat: 로그인 기능 추가"
IF style == SEMANTIC AND language == ENGLISH:
-> "feat: add login feature"
IF style == PLAIN AND language == KOREAN:
-> "로그인 기능 추가"
IF style == PLAIN AND language == ENGLISH:
-> "Add login feature"
IF style == SHORT:
-> "format" / "type fix" / "lint"
VALIDATION before each commit:
If ANY check fails -> REWRITE message.
</execution>
---
## PHASE 6: Verification & Cleanup
<verification>
### 6.1 Post-Commit Verification
```bash
# Check working directory clean
git status
# Review new history
git log --oneline $(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)..HEAD
# Verify each commit is atomic
# (mentally check: can each be reverted independently?)
IF fixup was used AND branch has upstream:
-> Requires: git push --force-with-lease
-> WARN user about force push implications
IF only new commits:
-> Regular: git push
COMMIT SUMMARY:
Strategy: <what was done>
Commits created: N
Fixups merged: M
HISTORY:
<hash1> <message1>
<hash2> <message2>
...
NEXT STEPS:
- git push [--force-with-lease]
- Create PR if ready
| If git log shows... | Use this style |
|---|---|
feat: xxx, fix: yyy | SEMANTIC |
Add xxx, Fix yyy, xxx 추가 | PLAIN |
format, lint, typo | SHORT |
| Full sentences | SENTENCE |
| Mix of above | Use MAJORITY (not semantic by default) |
Is this on main/master?
YES -> NEW_COMMITS_ONLY, never rewrite
NO -> Continue
Are all commits local (not pushed)?
YES -> AGGRESSIVE_REWRITE allowed
NO -> CAREFUL_REWRITE (warn on force push)
Does change complement existing commit?
YES -> FIXUP to that commit
NO -> NEW COMMIT
Is history messy?
YES + all local -> Consider RESET_REBUILD
NO -> Normal flow
STOP AND VERIFY - Do not proceed until ALL boxes checked:
[] File count check: N files -> at least ceil(N/3) commits?
- 3 files -> min 1 commit
- 5 files -> min 2 commits
- 10 files -> min 4 commits
- 20 files -> min 7 commits
[] Justification check: For each commit with 3+ files, did I write WHY?
[] Directory split check: Different directories -> different commits?
[] Test pairing check: Each test with its implementation?
[] Dependency order check: Foundations before dependents?
HARD STOP CONDITIONS:
<rebase_context>
# Execute ALL 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
| Condition | Risk Level | 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 | Will require force-push; confirm with user |
| All commits local | SAFE | Proceed freely |
| Upstream diverged | WARNING | May need --onto strategy |
USER REQUEST -> STRATEGY:
"squash commits" / "cleanup" / "정리"
-> INTERACTIVE_SQUASH
"rebase on main" / "update branch" / "메인에 리베이스"
-> REBASE_ONTO_BASE
"autosquash" / "apply fixups"
-> AUTOSQUASH
"reorder commits" / "커밋 순서"
-> INTERACTIVE_REORDER
"split commit" / "커밋 분리"
-> INTERACTIVE_EDIT
</rebase_context>
<rebase_execution>
# Find merge-base
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
# Start interactive rebase
# NOTE: Cannot use -i interactively. Use GIT_SEQUENCE_EDITOR for automation.
# For SQUASH (combine all into one):
git reset --soft $MERGE_BASE
git commit -m "Combined: <summarize all changes>"
# For SELECTIVE SQUASH (keep some, squash others):
# Use fixup approach - mark commits to squash, then autosquash
# When you have fixup! or squash! commits:
MERGE_BASE=$(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)
GIT_SEQUENCE_EDITOR=: git rebase -i --autosquash $MERGE_BASE
# The GIT_SEQUENCE_EDITOR=: trick auto-accepts the rebase todo
# Fixup commits automatically merge into their targets
# Scenario: Your branch is behind main, need to update
# Simple rebase onto main:
git fetch origin
git rebase origin/main
# Complex: Move commits to different base
# git rebase --onto <newbase> <oldbase> <branch>
git rebase --onto origin/main $(git merge-base HEAD origin/main) HEAD
CONFLICT DETECTED -> WORKFLOW:
1. Identify conflicting files:
git status | grep "both modified"
2. For each conflict:
- Read the file
- Understand both versions (HEAD vs incoming)
- Resolve by editing file
- Remove conflict markers (<<<<, ====, >>>>)
3. Stage resolved files:
git add <resolved-file>
4. Continue rebase:
git rebase --continue
5. If stuck or confused:
git rebase --abort # Safe rollback
| Situation | Command | Notes |
|---|---|---|
| Rebase going wrong | git rebase --abort | Returns to pre-rebase state |
| Need original commits | git reflog -> git reset --hard <hash> | Reflog keeps 90 days |
| Accidentally force-pushed | git reflog -> coordinate with team | May need to notify others |
| Lost commits after rebase | git fsck --lost-found | Nuclear option |
| </rebase_execution> |
<rebase_verify>
# Verify clean state
git status
# Check new history
git log --oneline $(git merge-base HEAD main 2>/dev/null || git merge-base HEAD master)..HEAD
# Verify code still works (if tests exist)
# Run project-specific test command
# Compare with pre-rebase if needed
git diff ORIG_HEAD..HEAD --stat
IF branch never pushed:
-> git push -u origin <branch>
IF branch already pushed:
-> git push --force-with-lease origin <branch>
-> ALWAYS use --force-with-lease (not --force)
-> Prevents overwriting others' work
</rebase_verify>
REBASE SUMMARY:
Strategy: <SQUASH | AUTOSQUASH | ONTO | REORDER>
Commits before: N
Commits after: M
Conflicts resolved: K
HISTORY (after rebase):
<hash1> <message1>
<hash2> <message2>
NEXT STEPS:
- git push --force-with-lease origin <branch>
- Review changes before merge
<history_search_type>
| User Request | Search Type | Tool |
|---|---|---|
| "when was X added" / "X가 언제 추가됐어" | PICKAXE | git log -S |
| "find commits changing X pattern" | REGEX | git log -G |
| "who wrote this line" / "이 줄 누가 썼어" | BLAME | git blame |
| "when did bug start" / "버그 언제 생겼어" | BISECT | git bisect |
| "history of file" / "파일 히스토리" | FILE_LOG | git log -- path |
| "find deleted code" / "삭제된 코드 찾기" | PICKAXE_ALL | git log -S --all |
From user request, identify:
- SEARCH_TERM: The string/pattern to find
- FILE_SCOPE: Specific file(s) or entire repo
- TIME_RANGE: All time or specific period
- BRANCH_SCOPE: Current branch or --all branches
</history_search_type>
<history_search_exec>
Purpose: Find commits that ADD or REMOVE a specific string
# Basic: Find when string was added/removed
git log -S "searchString" --oneline
# With context (see the actual changes):
git log -S "searchString" -p
# In specific file:
git log -S "searchString" -- path/to/file.py
# Across all branches (find deleted code):
git log -S "searchString" --all --oneline
# With date range:
git log -S "searchString" --since="2024-01-01" --oneline
# Case insensitive:
git log -S "searchstring" -i --oneline
Example Use Cases:
# When was this function added?
git log -S "def calculate_discount" --oneline
# When was this constant removed?
git log -S "MAX_RETRY_COUNT" --all --oneline
# Find who introduced a bug pattern
git log -S "== None" -- "*.py" --oneline # Should be "is None"
Purpose: Find commits where diff MATCHES a regex pattern
# Find commits touching lines matching pattern
git log -G "pattern.*regex" --oneline
# Find function definition changes
git log -G "def\s+my_function" --oneline -p
# Find import changes
git log -G "^import\s+requests" -- "*.py" --oneline
# Find TODO additions/removals
git log -G "TODO|FIXME|HACK" --oneline
-S vs -G Difference:
-S "foo": Finds commits where COUNT of "foo" changed
-G "foo": Finds commits where DIFF contains "foo"
Use -S for: "when was X added/removed"
Use -G for: "what commits touched lines containing X"
Purpose: Line-by-line attribution
# Basic blame
git blame path/to/file.py
# Specific line range
git blame -L 10,20 path/to/file.py
# Show original commit (ignoring moves/copies)
git blame -C path/to/file.py
# Ignore whitespace changes
git blame -w path/to/file.py
# Show email instead of name
git blame -e path/to/file.py
# Output format for parsing
git blame --porcelain path/to/file.py
Reading Blame Output:
^abc1234 (Author Name 2024-01-15 10:30:00 +0900 42) code_line_here
| | | | +-- Line content
| | | +-- Line number
| | +-- Timestamp
| +-- Author
+-- Commit hash (^ means initial commit)
Purpose: Find exact commit that introduced a bug
# Start bisect session
git bisect start
# Mark current (bad) state
git bisect bad
# Mark known good commit (e.g., last release)
git bisect good v1.0.0
# Git checkouts middle commit. Test it, then:
git bisect good # if this commit is OK
git bisect bad # if this commit has the bug
# Repeat until git finds the culprit commit
# Git will output: "abc1234 is the first bad commit"
# When done, return to original state
git bisect reset
Automated Bisect (with test script):
# If you have a test that fails on bug:
git bisect start
git bisect bad HEAD
git bisect good v1.0.0
git bisect run pytest tests/test_specific.py
# Git runs test on each commit automatically
# Exits 0 = good, exits 1-127 = bad, exits 125 = skip
# Full history of a file
git log --oneline -- path/to/file.py
# Follow file across renames
git log --follow --oneline -- path/to/file.py
# Show actual changes
git log -p -- path/to/file.py
# Files that no longer exist
git log --all --full-history -- "**/deleted_file.py"
# Who changed file most
git shortlog -sn -- path/to/file.py
</history_search_exec>
<history_results>
SEARCH QUERY: "<what user asked>"
SEARCH TYPE: <PICKAXE | REGEX | BLAME | BISECT | FILE_LOG>
COMMAND USED: 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 COMMIT: abc1234
DETAILS:
Author: John Doe <john@example.com>
Date: 2024-06-15
Files changed: 3
DIFF EXCERPT (if applicable):
+ def calculate_discount(price, rate):
+ return price * (1 - rate)
Based on search results, offer relevant follow-ups:
FOUND THAT commit abc1234 introduced the change.
POTENTIAL ACTIONS:
- View full commit: git show abc1234
- Revert this commit: git revert abc1234
- See related commits: git log --ancestry-path abc1234..HEAD
- Cherry-pick to another branch: git cherry-pick abc1234
</history_results>
| Goal | Command |
|---|---|
| When was "X" added? | git log -S "X" --oneline |
| When was "X" removed? | git log -S "X" --all --oneline |
| What commits touched "X"? | git log -G "X" --oneline |
| Who wrote line N? | git blame -L N,N file.py |
| When did bug start? | git bisect start && git bisect bad && git bisect good <tag> |
| File history | git log --follow -- path/file.py |
| Find deleted file | git log --all --full-history -- "**/filename" |
| Author stats for file | git shortlog -sn -- path/file.py |
--force instead of --force-with-lease -> DANGEROUS-S when -G is appropriate -> Wrong results-C on moved code -> Wrong attributionDesigner-turned-developer who crafts stunning UI/UX even without design mockups
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use only when the user explicitly requests agent-browser or browser automation is unavoidable. Avoid on Windows by default; do not use for routine post-work verification when non-browser checks can prove behavior.
Full PR lifecycle: git worktree → implement → atomic commits → PR creation → verification loop (CI + review-work + Cubic approval) → merge. Keeps iterating until ALL gates pass and PR is merged. Worktree auto-cleanup after merge. Use whenever implementation work needs to land as a PR. Triggers: 'create a PR', 'implement and PR', 'work on this and make a PR', 'implement issue', 'land this as a PR', 'work-with-pr', 'PR workflow', 'implement end to end', even when user just says 'implement X' if the context implies PR delivery.