fix-github-issue
Skill2 starsChanged 12 days agoMITView on GitHub
Fix a GitHub issue with quality gates — coverage, code review, E2E, and bugfix triage loop.
- Commits and pushes
5 other projects use a version of this.
Adopted by
Most recently taken up by
- simiancraft/simiancraft-skillstheir own version, first seen Nov 1, 2026
- drvoss/everything-copilot-clitheir own version, first seen Oct 26, 2026
- open-guji/luatex-cntheir own version, first seen Oct 9, 2026
- Shopify/flash-listtheir own version, first seen Mar 24, 2026
- AgnosticUI/agnosticuitheir own version, first seen Mar 2, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- Shopify/flash-list★ 7.2ksame name
- AgnosticUI/agnosticui★ 833same name
- open-guji/luatex-cn★ 127same name
- drvoss/everything-copilot-cli★ 48same name
- simiancraft/simiancraft-skills★ 8same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- claude-docsSkill
- create-agentSkill
- create-commandSkill
- create-skillSkill
- demoSkill
- design-e2eSkill
- execute-e2e-testsSkill
- telegramSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
17 rules. 0 are common, 11 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
11
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
If URL: extract number from path
- Very specific: line 25
If next: run gh issue list --label bug --state open --limit 1 --json number,title -q '.[0]' to get the highest priority open bug
- Distinctive: line 26
If number: use directly
- Distinctive: line 28
Run these validation checks directly:
- Distinctive: line 41
STOP if working tree is dirty or not on main branch.
- Distinctive: line 61
Search for relevant files using keywords from the issue
- Very specific: line 62
Check for linked PRs or related issues: gh issue list --search "mentions:#$ISSUE_NUMBER"
- Very specific: line 175
If COVERAGE_STATUS: WARN — log a warning but continue.
- Distinctive: line 176
If the coverage agent fails entirely — proceed to Phase 9 (bugfix loop).
- Very specific: line 189
If APPROVAL_STATUS: CHANGES_NEEDED persists after the review agent's fixes, proceed to Phase 9 (bugfix loop).
- Distinctive: line 204
If FAIL and not continuing, proceed to Phase 9 (bugfix loop).
- Very specific: line 217
If FIX_STATUS: FIXED and TESTS_PASSING: true:
- Distinctive: line 220
If build failed -> re-run Phase 5
- Distinctive: line 221
If coverage failed -> re-run Phase 6
- Distinctive: line 222
If review flagged issues -> re-run Phase 7
- Distinctive: line 223
If E2E failed -> re-run Phase 8
- Very specific: line 226
If --auto: log failure and STOP
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
DEPRECATED: Use
/fix-issueinstead. This skill is retained for reference but will be removed in a future update.
You are a thin dispatcher orchestrator. You delegate ALL heavy work to sub-agents and keep only argument parsing, control flow, and structured result parsing in your own context.
Phase 1: Parse Arguments & Validate Environment (DIRECT)
Parse $ARGUMENTS for:
- Target: issue number (e.g.,
123), issue URL (e.g.,https://github.com/owner/repo/issues/123), ornext(highest priority open bug) - Flags:
--skip-coverage— bypass coverage gate (build agent creates PR directly)--e2e-gate=block|warn|off— E2E test gate behavior (default:off)--auto— skip interactive prompts on failure
Extract the issue number from the argument:
- If URL: extract number from path
- If
next: rungh issue list --label bug --state open --limit 1 --json number,title -q '.[0]'to get the highest priority open bug - If number: use directly
Run these validation checks directly:
# 1. GitHub CLI auth
gh auth status
# 2. Clean git state
git status --porcelain
# 3. On main branch
git branch --show-current
# 4. Pull latest
git pullSTOP if working tree is dirty or not on main branch.
Phase 2: Fetch Issue & Validate (DIRECT)
gh issue view $ISSUE_NUMBER --json number,title,body,state,assignees,labelsStop conditions — STOP and report if:
- Issue is
closed - Issue is assigned to someone else (not the current
ghuser) - Issue has label
wontfixorwon't fix
Extract: ISSUE_NUMBER, ISSUE_TITLE, ISSUE_BODY, ISSUE_LABELS.
Phase 3: Investigation (DIRECT)
Investigate the issue yourself (this is lightweight context gathering):
- Extract key details: reproduction steps, error messages, stack traces, affected files
- Search for relevant files using keywords from the issue
- Check for linked PRs or related issues:
gh issue list --search "mentions:#$ISSUE_NUMBER" - Review recent commits touching affected files
- Assess complexity (simple / moderate / complex)
Phase 4: Fix Plan & Approval Gate (DIRECT)
Present a structured plan:
### Proposed Fix for #[ISSUE_NUMBER]: [ISSUE_TITLE]
**Root Cause**: [your diagnosis]
**Approach**: [strategy in 1-2 sentences]
**Complexity**: Simple / Moderate / Complex
**Files to Modify**:
- `path/to/file` — [what changes]
**New Files** (if any):
- `path/to/new` — [purpose]
**Tests**:
- [ ] Update existing: [which]
- [ ] Add new: [describe coverage]
**Risk Assessment**: Low / Medium / High + rationale
**Quality Gates**:
- Coverage gate: [enabled / skipped]
- E2E gate: [off / warn / block]STOP and wait for user approval before proceeding.
Phase 5: Build Agent
Launch the appropriate agent (detect from project type — backend-typescript-architect for TS/Bun, python-backend-engineer for Python, general-purpose as fallback).
If --skip-coverage is set (build agent handles push + PR):
You are fixing GitHub issue #[ISSUE_NUMBER]: [ISSUE_TITLE]
## Issue Details
[ISSUE_BODY]
## Fix Plan
[The approved plan from Phase 4]
## Instructions
1. Create branch: git checkout -b fix/issue-[ISSUE_NUMBER]-[short-description]
2. Reproduce the bug first — create a failing test that demonstrates the issue
3. Implement the minimal fix
4. Add defensive tests — edge cases, regression prevention
5. Run all quality gates (tests, types, lint)
6. Commit:
git add [specific files]
git commit -m "fix: [description]
Fixes #[ISSUE_NUMBER]
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>"
7. Push and create PR:
git push -u origin fix/issue-[ISSUE_NUMBER]-[short-description]
gh pr create --title "fix: [ISSUE_TITLE] (#[ISSUE_NUMBER])" --body "[summary, testing, Fixes #[ISSUE_NUMBER]]"
Return PR_NUMBER: [number] and PR_URL: [url] when done.Extract PR_NUMBER from the agent result. Skip Phase 6.
If coverage gate is enabled (default — build agent commits locally only):
You are fixing GitHub issue #[ISSUE_NUMBER]: [ISSUE_TITLE]
## Issue Details
[ISSUE_BODY]
## Fix Plan
[The approved plan from Phase 4]
## Instructions
1. Create branch: git checkout -b fix/issue-[ISSUE_NUMBER]-[short-description]
2. Reproduce the bug first — create a failing test that demonstrates the issue
3. Implement the minimal fix
4. Add defensive tests — edge cases, regression prevention
5. Run all quality gates (tests, types, lint)
6. Commit locally:
git add [specific files]
git commit -m "fix: [description]
Fixes #[ISSUE_NUMBER]
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>"
7. DO NOT push or create a PR — the coverage agent handles that next.
Return BRANCH_NAME: fix/issue-[ISSUE_NUMBER]-[short-description] and BUILD_STATUS: SUCCESS when done.Extract BRANCH_NAME and BUILD_STATUS from the agent result. Proceed to Phase 6.
Phase 6: Coverage Gate (skip if --skip-coverage)
Launch a qa-engineer agent with the prompt from ${CLAUDE_SKILL_DIR}/coverage-gate-prompt.md, substituting:
{{ISSUE_NUMBER}}-> current issue number{{ISSUE_TITLE}}-> current issue title{{BRANCH_NAME}}-> branch name from build agent result
The agent runs coverage analysis, adds tests, pushes the branch, and creates the PR.
Extract PR_NUMBER, PR_URL, COVERAGE_PCT, TESTS_ADDED, and COVERAGE_STATUS from the agent result.
If COVERAGE_STATUS: WARN — log a warning but continue.
If the coverage agent fails entirely — proceed to Phase 9 (bugfix loop).
Phase 7: Review Gate
Launch a senior-code-reviewer agent with the prompt from ${CLAUDE_SKILL_DIR}/review-gate-prompt.md, substituting:
{{ISSUE_NUMBER}}-> current issue number{{ISSUE_TITLE}}-> current issue title{{PR_NUMBER}}-> PR number from Phase 5 (if--skip-coverage) or Phase 6 (default)
The agent reviews the PR, fixes issues if found, and approves when satisfied.
Extract APPROVAL_STATUS from the agent result.
If APPROVAL_STATUS: CHANGES_NEEDED persists after the review agent's fixes, proceed to Phase 9 (bugfix loop).
Phase 8: E2E Gate (skip unless --e2e-gate=block|warn)
Launch a qa-engineer agent with the prompt from ${CLAUDE_SKILL_DIR}/e2e-gate-prompt.md, substituting:
{{ISSUE_NUMBER}}-> current issue number{{ISSUE_TITLE}}-> current issue title{{PR_NUMBER}}-> PR number{{BRANCH_NAME}}-> branch name
Handle result per --e2e-gate mode:
block+ FAIL: if--autotreat aswarn, otherwise ask user: retry / continue / abortwarn+ FAIL: log warning, continue- PASS: continue
If FAIL and not continuing, proceed to Phase 9 (bugfix loop).
Phase 9: Bugfix Loop (on any gate failure)
Launch a general-purpose agent with the prompt from ${CLAUDE_SKILL_DIR}/bugfix-agent-prompt.md, substituting:
{{ISSUE_NUMBER}}-> current issue number{{ISSUE_TITLE}}-> current issue title{{BRANCH_NAME}}-> the fix branch{{FAILED_STEP}}-> which phase failed (build | coverage | review | e2e){{FAILURE_OUTPUT}}-> the error/failure text from the failed agent
Extract FAILURE_CATEGORY, ISSUE_NUMBER (sub-issue), FIX_STATUS, TESTS_PASSING from the agent result.
- If
FIX_STATUS: FIXEDandTESTS_PASSING: true:- Log: "Fixed ([FAILURE_CATEGORY]) — retrying from failed step"
- Re-run from the phase that failed (not from Phase 5):
- If build failed -> re-run Phase 5
- If coverage failed -> re-run Phase 6
- If review flagged issues -> re-run Phase 7
- If E2E failed -> re-run Phase 8
- Allow max 2 bugfix iterations to prevent infinite loops
- If
FIX_STATUS: UNFIXED:- If
--auto: log failure and STOP - Otherwise: ask user what to do
- If
Phase 10: Merge & Cleanup (DIRECT)
Once all gates pass:
# Merge PR
gh pr merge $PR_NUMBER --squash --delete-branch
# Comment on and close original issue
gh issue comment $ISSUE_NUMBER --body "Fixed in PR #$PR_NUMBER."
gh issue close $ISSUE_NUMBER --reason completed
# Return to main
git checkout main
git pullPhase 11: Summary Output (DIRECT)
Print a structured summary:
## Fix Complete: #[ISSUE_NUMBER] — [ISSUE_TITLE]
- **PR**: #[PR_NUMBER] ([PR_URL])
- **Branch**: [BRANCH_NAME] (merged & deleted)
- **Files Modified**: [count]
- **Tests Added**: [count]
- **Coverage**: [COVERAGE_PCT]% [COVERAGE_STATUS]
- **Review**: [APPROVAL_STATUS]
- **E2E**: [E2E_RESULT or "skipped"]
- **Bugfix Iterations**: [count]Context Budget Rules
NEVER read project source files directly — the build agents read them. Your per-phase context cost should be minimal — delegate all heavy lifting to sub-agents.
Context
$ARGUMENTS
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

