You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Adopt ACF in homedir to test the stack-audit and label-metadata phases in a real autonomous SDLC pipeline. homedir already has issue-metadata-validation CI, scc-* labels, and an autonomous worker — ACF should complement, not replace, these systems.
700+ issues and 1200+ PRs of historical data to audit
The question isn't "does ACF work in theory" but "does ACF's stack-audit find orphan PRs and escaped issues that homedir's existing systems miss?" If ACF finds gaps that the existing CI doesn't catch, that's immediate value. If it finds the same gaps, ACF is redundant. If it finds nothing, ACF's stack-audit is broken.
From an AI engineering perspective (30+ years watching systems fail): the only honest test is against a system that already has its own checks. Testing against a greenfield project tells you nothing — everything works on empty repos.
Affected Files (in homedir — tracked here, executed there)
homedir/.devin/skills/ or homedir/.opencode/skills/ — ACF installation
Constraint: This issue lives in the ACF repo. All file changes happen in homedir via separate PRs there. This issue tracks the adoption plan and results.
Acceptance Criteria
ACF installed in homedir via ./install.sh /home/vector/Proyectos/homedir --agent opencode
ACF canonical labels created in homedir (if not already present):
Verify these don't conflict with existing scc-* labels
Stack-audit run on homedir:
gh pr list --state open → check for orphan PRs (no Closes #N)
gh issue list --state open → check for stale issues (ready-to-implement, no PR)
gh pr list --state merged --limit 50 → check for close gaps (merged PR, issue still open)
gh issue list --state closed --limit 50 → check for escaped issues (closed, no merged PR)
Results compared against homedir's existing issue-metadata-validation CI:
Does ACF find gaps that the CI doesn't catch?
Does ACF's label taxonomy align with or conflict with scc-* labels?
Is ACF's stack-audit output more or less useful than the CI output?
Findings documented in a comment on this issue:
What ACF found
What homedir's existing systems already caught
Where ACF adds value vs redundancy
Recommended adjustments to ACF based on homedir's patterns
If ACF's label taxonomy conflicts with homedir's scc-* labels, document the conflict and propose a resolution (e.g., ACF labels as a superset, or a mapping table)
ACF stack-audit output is non-empty (it found something to report)
homedir git status is clean after ACF installation (ACF is additive)
Complexity
Medium — installation is trivial (./install.sh), but the audit comparison requires understanding homedir's existing CI and label system.
Notes
homedir's issue-metadata-validation.yml already checks for required labels on issue create/edit. ACF's stack-audit is broader: it checks the relationship between issues and PRs, not just issue metadata.
homedir's autonomous-implementation.md template is very similar to ACF's issue-contextualized.md template — compare them and document differences.
The scc-* labels (scc-ready, scc-failing-checks, etc.) are homedir-specific. ACF's labels (priority:P0-P3, area:*) are generic. They should coexist.
This is a Phase 2 (Beta) issue per the roadmap — real-world testing.
Summary
Adopt ACF in homedir to test the stack-audit and label-metadata phases in a real autonomous SDLC pipeline. homedir already has issue-metadata-validation CI, scc-* labels, and an autonomous worker — ACF should complement, not replace, these systems.
Context
ARCHITECTURE.md(Quarkus/Go dual-stack).agent/skills/homedir_expert/,.opencode/skills/wos-cycle/,.opencode/skills/wos-delegation/.github/workflows/issue-metadata-validation.yml(validates issue labels on create/edit).github/ISSUE_TEMPLATE/autonomous-implementation.mdskills/02-stack-audit/SKILL.mdskills/06-label-metadata/SKILL.mdMotivation
homedir is the ideal testbed for ACF's stack-audit because it already has:
The question isn't "does ACF work in theory" but "does ACF's stack-audit find orphan PRs and escaped issues that homedir's existing systems miss?" If ACF finds gaps that the existing CI doesn't catch, that's immediate value. If it finds the same gaps, ACF is redundant. If it finds nothing, ACF's stack-audit is broken.
From an AI engineering perspective (30+ years watching systems fail): the only honest test is against a system that already has its own checks. Testing against a greenfield project tells you nothing — everything works on empty repos.
Affected Files (in homedir — tracked here, executed there)
homedir/.devin/skills/orhomedir/.opencode/skills/— ACF installationhomedir/.github/workflows/— potential ACF-triggered workflowAcceptance Criteria
./install.sh /home/vector/Proyectos/homedir --agent opencodearea:backend,area:frontend,area:ci,area:docs,area:security,area:devopsgh pr list --state open→ check for orphan PRs (noCloses #N)gh issue list --state open→ check for stale issues (ready-to-implement, no PR)gh pr list --state merged --limit 50→ check for close gaps (merged PR, issue still open)gh issue list --state closed --limit 50→ check for escaped issues (closed, no merged PR)Validation
gh issue list --repo VECTORG99/homedir --label "area:backend"returns results (labels exist)git statusis clean after ACF installation (ACF is additive)Complexity
Medium — installation is trivial (
./install.sh), but the audit comparison requires understanding homedir's existing CI and label system.Notes
issue-metadata-validation.ymlalready checks for required labels on issue create/edit. ACF's stack-audit is broader: it checks the relationship between issues and PRs, not just issue metadata.autonomous-implementation.mdtemplate is very similar to ACF'sissue-contextualized.mdtemplate — compare them and document differences.