-
Notifications
You must be signed in to change notification settings - Fork 0
Open
Labels
enhancementNew feature or requestNew feature or request
Description
Goal
Full audit of every gh-aw workflow in this repo against the official gh-aw documentation (https://github.github.com/gh-aw/) and samples in github.com/github/gh-aw. We have been learning gh-aw through trial and error and have hit multiple issues that could have been avoided by reading the docs first. This audit should catch any remaining misconfigurations before they bite us in production.
Scope
For each gh-aw workflow (.md + .lock.yml):
1. Triggers
- Is the trigger type the best fit for our use case?
- Are we using
workflow_dispatchinputs correctly? - Do we need
if:frontmatter conditions? Are they correct? - Are
roles:andbots:configured properly (or omitted due to gh-aw#21098)?
2. Safe Outputs
- Is each safe output
targetvalue correct for the trigger type?"triggering"for event-based triggers${{ inputs.pr_number }}forworkflow_dispatchwhen the handler lacks a PR number field"*"forworkflow_dispatchwhen the handler hasitem_number/comment_idin its schema
- Are
maxlimits appropriate? - Are
github-tokenvalues set where needed? - Are we missing any useful safe outputs? (e.g.,
update-pull-requestfor Responder cannot update PR title/description — add update_pull_request safe output #165,resolve-pull-request-review-thread) - Are
footer,labels,protected-filesconfigs optimal?
3. Permissions
- Are we requesting minimum necessary permissions?
- Any permissions we have that we dont need?
- Any permissions we need that we dont have?
4. Network
- Are
network.alloweddomains correct? (e.g.,astral.shkeeps getting blocked for uv/ruff installs) - Are we allowing too much?
5. Engine / Model
- Are we using the right model for each workflow?
- Is Opus overkill for any of these? Could Sonnet handle some?
6. Tools
- Are
toolsetsconfigured correctly? - Are we using
--allow-all-toolswhere we should/shouldnt?
7. Checkout / Fetch
- Do workflows that push code have
fetch: ["*"]andfetch-depth: 0? - Do read-only workflows unnecessarily fetch too much?
Workflows to audit
| Workflow | File | Current trigger |
|---|---|---|
| Issue Implementer | issue-implementer.md |
workflow_dispatch |
| Review Responder | review-responder.md |
workflow_dispatch |
| CI Fixer | ci-fixer.md |
workflow_dispatch |
| Quality Gate | quality-gate.md |
workflow_dispatch |
| Code Health | code-health.md |
TBD |
| Test Analysis | test-analysis.md |
TBD |
| Pipeline Orchestrator | pipeline-orchestrator.yml |
workflow_run / push / pull_request_review / workflow_dispatch |
Reference
- gh-aw docs: https://github.github.com/gh-aw/
- Safe outputs reference: https://github.github.com/gh-aw/reference/safe-outputs/
- Samples: https://github.com/github/gh-aw (if accessible)
- Related issues: Responder cannot update PR title/description — add update_pull_request safe output #165 (missing update-pull-request), Quality gate never fires on clean PRs — orchestrator missing happy path #168 (quality gate dispatch)
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
enhancementNew feature or requestNew feature or request