| name | plan-verification |
| description | Use when an FSD and goal issue board need requirement coverage, goal quality, dependency DAG, sizing, and verification validated before execution. |
Plan Verification
Announce: "I'm using the plan-verification skill to validate this plan before execution."
Reference Router
Load only the dimensions involved in the current check or failed re-check:
- Requirement coverage, task completeness, and dependency DAGs: coverage and dependencies
- Key links, scope sanity, and derived must-haves: links, scope, and must-haves
- Goal and issue granularity: sizing
- Critical-path tests and exact technical-decision coverage: tests and decisions
- For UI-bearing scope, load the canonical profile, score, and hard gates from UI contract readiness.
- Severity, report format, verdict, and retry limit: verification process
- Permitted targeted fixes: revision rules
Run all ten verification dimensions initially; after revision rerun failed dimensions and affected dependencies.
Mandatory Gates
- Coverage gate: Every BRD/PRD requirement and acceptance criterion needs corresponding FSD authority, verification, and a goal issue pointer. Missing coverage is Critical.
- Goal gate: Every goal needs a clear action, verification, done criteria, explicit dependencies, and a coherent independently verifiable outcome. Dependencies and
Blocked by paths must exist, precede dependents, and form an acyclic graph.
- Link gate: Database, API, UI, shared-type, workflow, and other key connections must be ordered and specified before consumers rely on them.
- Scope gate: Derive endpoint/page/workflow capability, critical error handling, and user validation. Split separate features or oversized work; merge tiny mechanical work. A coherent tracer bullet may cross layers and must not be rejected merely for doing so.
- Test gate: Every critical path, edge case, and error path needs a verification step.
- Decision gate: Every approved
TDEC-* and every applicable obligation from a linked ACCEPTED ADR must map by exact ID to at least one GOAL-* and one TEST-*. Blocked or superseded decisions stay out of executable goals. Missing exact-ID coverage blocks execution. Fuzzy text similarity may warn after implementation but must never create a blocking match.
- UI/API readiness gate: Dimension 10 applies only to UI-bearing scope. Require
score at least 90 plus every canonical hard gate, contract version and refs on
each UI issue, exactly one first vertical slice, and scale-out dependencies on
its verified issue. Require exactly one
HARDENING goal that depends on every
UI delivery slice. A numeric score, exception, or mock-only result cannot hide
a failed hard gate. Non-UI work records NOT_APPLICABLE with approved reason.
- Enabler-only gate: A
CONTRACT_ENABLER is the sole exception while
readiness is DRAFT or BLOCKED. Report PASS WITH NOTES - ENABLER_ONLY only
when the enabler is independently bounded/testable and every
first vertical slice and scale-out issue remains blocked. After verification, return to
/sc-plan and rerun all UI/API hard gates.
- Verdict gate: Report
PASS, PASS WITH NOTES, or NEEDS REVISION, with findings classified Critical, Important, or Suggestion. Apply targeted fixes only; do not change scope or add features.
- Stop gate: After three failed revision loops, stop and mark
needs_review.