stack-audit
Skill2 starsChanged 3 months agoMITView on GitHub
Phase 2 of ACF. Audits the open GitHub stack for orphan PRs, unclosed issues, missing issue↔PR references, and library opportunities. Use when ACF delegates stack analysis, or when the user asks to "check the stack", "find orphan PRs", "review open issues", or "what's missing in the stack".
2 other projects use a version of this.
Adopted by
Most recently taken up by
- truemetal74/skillstheir own version, first seen Oct 22, 2026
- patmeh1/product-docs-kittheir own version, first seen Sep 9, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- patmeh1/product-docs-kit★ 0same name
- truemetal74/skills★ 0same 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
- context-loadSkill
- issue-craftSkill
- pr-contextSkill
- frontend-previewSkill
- label-metadataSkill
- compactionSkill
- cavemanSkill
- graph-scopeSkill
- acfSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 3 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 77
Do NOT bundle it into the current issue
- Very specific: line 78
Create a separate enhancement issue with labels enhancement + library-review
- Distinctive: line 90
Write findings to the progress file:
- Very specific: line 120
Use --limit 200 or higher if the API allows it
- Distinctive: line 129
Document in the findings if pagination was used and whether all pages were consumed
- Very specific: line 150
ghnot authenticated: run gh auth status to confirm. - Very specific: line 153
GitHub API rate limits: if a gh command returns a rate limit error, stop and report "GitHub API rate limited — retry after [reset time].
- Very specific: line 160
Commands fail: if a gh command fails (network error, permission denied), record the error in findings as [check name]: failed — [error …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Stack-Audit
Detects gaps in the issue↔PR pipeline so nothing escapes the current stack.
Checks
1. Orphan PRs (PR with no issue reference)
gh pr list --state open --json number,title,body,headRefNameFor each PR, check if the body contains Closes #N, Fixes #N, Refs #N, or
Resolves #N. If none found → orphan PR.
Flag: orphan-pr label suggestion + comment recommending issue creation.
2. Stale Issues (issue ready/in-progress with no PR)
gh issue list --state open --json number,title,labelsFor each issue with label ready-to-implement (or equivalent), check if there's
an open PR referencing it. If not → stale issue.
Flag: comment with "No PR open for this issue — consider creating one or closing if obsolete."
3. Merged PR, Issue Still Open
gh pr list --state merged --json number,title,body --limit 100For each merged PR with Closes #N, verify issue #N is closed. If open →
close gap.
Action: do not close automatically. Surface the gap in the findings list and let the user confirm. Only after explicit user approval run:
gh issue close #N --reason completedClosing an issue is a state change on the tracker; never perform it without user confirmation, and never batch-close multiple issues in one pass.
4. Closed Issue, No Merged PR
gh issue list --state closed --json number,title,body --limit 100For each closed issue, check if there's a merged PR referencing it. If not → escaped issue (closed without implementation evidence).
Flag: escaped-issue label suggestion. May need reopening or a follow-up issue.
5. Library Opportunities
During context-load and stack-audit, if a dependency or library could simplify the solution to an open issue:
- Do NOT bundle it into the current issue
- Create a separate enhancement issue with labels
enhancement+library-review - Reference it from the main issue:
Related: #NN (library review)
The library-review issue should contain:
- Library name and link
- What problem it solves
- Which open issues it could improve
- Trade-offs (license, bundle size, maintenance status)
Output
Write findings to the progress file:
## Stack-Audit Findings
### Orphan PRs
- PR #NN "title" — no issue reference
### Stale Issues
- Issue #NN "title" — no open PR
### Close Gaps
- PR #NN merged but issue #MM still open
### Escaped Issues
- Issue #NN closed but no merged PR
### Library Opportunities
- [library name] could simplify issue #NN — create enhancement issueParallel Execution
Stack-audit can run in parallel with context-load. Both feed into issue-craft.
Pagination
The default --limit is set to 100 (not 20) to cover most repos in a single
call. For very large repos with more than 100 open/merged items:
- Use
--limit 200or higher if the API allows it - Make multiple paginated calls with
--limit 100and track the last seen number to avoid duplicates - For targeted queries, use
gh search prs/gh search issueswith search terms instead of listing everything:gh search prs --state open --review-requested @me gh search issues --state open --label "ready-to-implement" - Document in the findings if pagination was used and whether all pages were consumed
Empty State Handling
- No open issues: report
Open issues: 0and note "No open issues — stack is clean or tracker is unused." Proceed to PR checks. - No open PRs: report
Open PRs: 0and note "No open PRs." Proceed to issue checks. - No issues AND no PRs: report "Empty stack — no issues or PRs found.
Repo may be new or
ghmay lack access." Suggest verifyinggh auth status. - No merged PRs: skip check 3 (close gaps). Note "No merged PRs — close gap check skipped."
- No closed issues: skip check 4 (escaped issues). Note "No closed issues — escaped issue check skipped."
Error Handling
ghCLI not installed: stack-audit cannot run withoutgh. Report "gh CLI not found — stack-audit skipped. Install: https://cli.github.com/" and continue with context-load only.ghnot authenticated: rungh auth statusto confirm. If not authed, report "gh not authenticated — rungh auth loginand retry." Do not attempt to authenticate automatically.- GitHub API rate limits: if a
ghcommand returns a rate limit error, stop and report "GitHub API rate limited — retry after [reset time]. Partial findings may be incomplete." Note which checks completed and which were skipped. - Required files/directories don't exist: stack-audit operates on GitHub
data, not local files. If the repo has no
.github/directory, that's expected — templates are checked by context-load, not stack-audit. - Commands fail: if a
ghcommand fails (network error, permission denied), record the error in findings as[check name]: failed — [error message]and continue with the next check. Do not abort the entire audit.
Example
Filled-in findings for a sample project:
## Stack-Audit Findings
### Orphan PRs
- PR #42 "feat: add leaderboard UI" — no issue reference
### Stale Issues
- Issue #17 "fix: flashlight uses :Fire() instead of :FireServer()" — no open PR
### Close Gaps
- PR #38 merged but issue #29 still open — awaiting user confirmation
### Escaped Issues
- Issue #51 closed but no merged PR — may need follow-up
### Library Opportunities
- Roact could simplify issue #17 (UI state management) — create enhancement issueDoes 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.

