| name | plan-validation |
| description | Judgment-based plan quality checks — cross-reference integrity, dependency ordering, contract completeness, terminology consistency |
Plan Validation Skill
Judgment checks that can't be scripted deterministically. Invoked by /meta-planner Stage 4.
Checks
-
Cross-reference integrity — Do master plan task IDs match phase file task IDs 1:1? Are any tasks in master missing from phase files or vice versa?
-
Dependency ordering — Does the task order respect declared dependencies? Are there tasks whose Depends on: references a task that appears later in the sequence?
-
Contract completeness — For every API endpoint in the design doc, is there a corresponding implementation task? For every UI component, is there a corresponding test?
-
Terminology consistency — Are naming conventions consistent across all files? PascalCase classes, Test prefix, project file naming.
-
File path accuracy — Do all file paths in the plan match actual project structure? Any stale paths from renames?
-
Checker completeness — Does every Verify-After command test what it claims to test? No verification commands that always pass.
-
Focused verification — Does every automated Verify-After name one test
file/node or explicitly scoped declared-file check? Reject bare/directory pytest,
-k without a file, package-wide npm/Vitest/Jest, npm run check,
svelte-check, project-wide tsc, builds, and full suites anywhere in the
execution plan, including phase-end acceptance. Broad release checks belong to
CI/ship, not /meta-execute. Critical-risk tasks must name a focused verifier
or an explicit human gate.
Output Format
## Plan Validation Report — <plan-name>
### Cross-Reference
- [ finding ]
### Dependency Ordering
- [ finding ]
### Contract Completeness
- [ finding ]
### Terminology
- [ finding ]
### File Paths
- [ finding ]
### Verification Soundness
- [ finding ]
### Verification Scope
- [ finding ]
## Summary
- Errors: N | Warnings: M
