| name | feature-finish |
| description | Graduation workflow for completing a feature. Reviews spec completion, updates durable product docs, enriches aiwiki/, captures deferrals as ideas, and verifies backlog. Does NOT commit or merge. |
Feature Finish
Run this skill after implementation is complete but before committing. It ensures that documentation, knowledge base, and backlog stay current with the code changes.
Process
1. Identify the spec
Find the active spec directory under specs/ for this feature. If there is no spec (ad-hoc work), note that and skip spec-related steps.
2. Diff summary
Review the working changes (git diff --stat and git diff --name-only) to understand what was touched. Summarize:
- Which layers were modified (routes, components, orchestrators, stores, utils, backend, lcc-rs)
- Which modules were added, changed, or removed
- Which tests were added or changed
3. Assess product/ impact
Check whether any user-visible behavior, workflow, or architecture ownership changed:
- product/glossary.md: Do any new terms need to be added? Were any existing terms redefined by this work?
- product/architecture/: Do code-placement rules or ownership boundaries need updating?
- product/architecture/adr/: Were architecture decisions made during this work that should be recorded?
Update the affected product/ files inline.
4. Assess aiwiki/ impact
Check whether the AI knowledge base needs updates for the touched modules:
- aiwiki/owners.md: Are all new or changed modules listed? Are test file mappings current? Are shared conventions updated if a new pattern was established?
- aiwiki/seams.md: For every seam the feature touched, confirm Contributors and Consumers are current. If the feature added either, the entry should have been updated during
/build post-implementation — confirm here as a safety net. If the feature introduced a new seam (a contract with one Owner and ≥2 Contributors or ≥2 Consumers where none existed before), add a new entry. Bump Last-audited if you re-greped during this audit.
- aiwiki/flows.md: Did any workflow's module participation change?
- aiwiki/architecture-health.md: Were any coupling risks, architecture debt items, or depth assessments affected?
Update the affected aiwiki/ files inline.
5. Consistency check
Verify no stale references:
- Grep for references to renamed or removed files in product/, aiwiki/, and .github/
- Check that any new shared logic added during the feature is documented in owners.md shared conventions section
- Check that integration boundaries in owners.md still match reality
6. Capture deferrals as ideas
Review conversations, TODOs-in-code, and spec notes for work that was deferred during implementation:
- For each deferred item, propose a GitHub issue (title,
kind/* plus relevant area/* labels, body using the idea template fields: Summary, Areas, Origin, Prior Work, Open Questions). Present the proposals to the user and wait for explicit confirmation before creating any issue. Do not pre-apply a status/* label.
- Do not create issues for trivial cleanup — only for items with enough substance to save future re-discovery.
7. Backlog check
Read specs/backlog.md and update it:
- Mark resolved items as done
- Add newly revealed backlog items
- Update items whose scope or priority changed due to this work
8. Spec status
If this work has an active spec:
- Review whether all spec requirements were addressed
- Note any requirements that were descoped or deferred (should already be captured as ideas in step 6)
9. Summary
Output a concise summary:
- What product/ docs were updated and why
- What aiwiki/ docs were updated and why
- What ideas were captured
- What backlog items were resolved or added
- Any open items that still need attention
Do NOT commit, merge, or push. The user will review all changes first.