Feature Audit
Ensures features are not just implemented but actually user-accessible. Bridges the gap between "code complete" and "user can use it."
When NOT to Use
- Analyzing agent/session patterns — Use
session-reviewfor retrospectives on how work went - Security review — Use
security-auditfor vulnerability scanning; this skill checks feature completeness, not security - Creating the plan — Use
planningto break down PRDs into tasks; this skill audits after implementation
Completeness Levels
| Level | Name | Meaning |
|---|---|---|
| 0 | Not Started | No implementation evidence |
| 1 | Backend Only | Service/API exists, no frontend |
| 2 | Frontend Exists | UI components exist, not accessible |
| 3 | Routable | Has route/screen, not in navigation |
| 4 | Accessible | In navigation, users can reach it |
| 5 | Complete | Accessible + documented + tested |
Quick Start Checklist
- Discover project framework and structure
- Load feature/task context from ohno (
mcp__ohno__get_epics,mcp__ohno__get_tasks) ornpx @stevestomp/ohno-cli tasks - For each feature: check backend, frontend, route, navigation
- Assign completeness level (0-5)
- Generate report and remediation tasks
Key Principles
- Always verify in codebase — don't trust task status alone
- Check full accessibility chain: backend → frontend → route → navigation → docs
- Generate specific remediation tasks (not generic "add frontend")
- Priority follows feature priority (P0 gap = P0 remediation)
References
| Reference | Description |
|---|---|
| framework-patterns.md | Scanning patterns for each framework |
| completeness-criteria.md | Detailed level definitions and checklists |
| gap-analysis.md | Analysis methodology |
| remediation-templates.md | Task templates for common gaps |
| scanning-process.md | Backend, frontend, API, navigation scans |
| anti-patterns.md | Audit and remediation anti-patterns |
Runtime Notes
- Claude Code: dispatch yokay-auditor via the Task tool with
subagent_type: "pokayokay:yokay-auditor". - Codex: there is no subagent dispatch. Execute the agent's role inline — read the corresponding
agents/yokay-<name>.mdand follow its Behavioral Defaults, Critical Rules, and Output Contract directly in the current session.