prd-audit
Audit a PRD against the current codebase before implementation. Verifies entity names, property names, and API signatures are still accurate. TRIGGER when the user says "audit the PRD", "check the PRD
SKILL.md.claude/skills/prd-audit/SKILL.md
GitHubPRD Freshness Audit
PRD to audit: $ARGUMENTS
If a PRD is >2 weeks old, it may reference names, properties, or APIs that have changed. This audit verifies accuracy before implementation.
Step 1: Read the PRD
Read the specified PRD file completely.
Step 2: API/Interface Verification
For each function, method, or API referenced in the PRD:
- Search the actual source file for the signature
- Verify parameter names, types, and return values match
- Check if the function still exists or was renamed/removed
Step 3: Entity Verification (Core Data)
For each Core Data entity referenced in the PRD:
- Search
Models/<EntityName>+CoreDataProperties.swiftfor actual property names and types - Compare with what the PRD describes
- Flag any mismatches
Step 4: File Structure Verification
For each file or module referenced in the PRD:
- Verify the file exists at the stated path
- Check if the structure matches the PRD description
- Note any new or removed files since the PRD was written
Step 5: Save Count Verification
Count the number of context.save() calls described in the PRD flow and compare with the actual service implementation. Mismatched save counts indicate architectural drift.
Step 6: Report
## PRD Audit: [PRD Name]
### PRD Age
- Created: [date]
- Last updated: [date]
- Age: [days]
### Entity Accuracy
- [Entity]: [MATCH / MISMATCH — details]
### API Accuracy
- [Module.function()]: [MATCH / MISMATCH — details]
### File Accuracy
- [FileName]: [MATCH / MISMATCH — details]
### Save Count Accuracy
- PRD describes: [N] saves
- Actual code: [N] saves
- [MATCH / MISMATCH]
### Recommended PRD Updates
- [List of specific corrections needed]