What This Skill Does

Validates a plan file against 8 dimensions to catch gaps before implementation begins. Does NOT replace Gate 1 (human approval) -- supplements it with systematic, automated checks. Every dimension has binary pass criteria; nothing is subjective.

When to Use

• After Gate 1 approval, before Phase 2 (Test)
• When /mk:cook detects a COMPLEX task (auto-suggested)
• User says "validate this plan", "check my plan", "is this plan complete"
• Before committing to implementation, to stress-test plan quality
• NOT for green-field harness sprints -- use mk:sprint-contract instead

Core Capabilities

1. 8-Dimension Check: Each dimension has explicit pass criteria and common failure patterns:
   • Scope Clarity -- in/out-of-scope explicit and non-overlapping
   • Acceptance Criteria -- every criterion binary (pass/fail), not subjective
   • Dependencies Resolved -- all external dependencies identified with status (available/blocked)
   • Risks Identified -- at least 1 risk flag with mitigation strategy
   • Architecture Documented -- technical approach references existing patterns or includes ADR
   • Test Strategy -- covers acceptance criteria; edge cases identified
   • Security Considered -- auth, data access, input validation addressed or explicitly N/A
   • Effort Estimated -- time/complexity estimate with confidence level
2. Whole-Plan Consistency Sweep: Before emitting the report, cross-checks plan.md and all phase-*.md files for stale references, renamed entities, scope changes, and dropped features. Reconciles contradictions across files
3. Routing: All 8 PASS -> proceed to Phase 2. Any FAIL -> return to planner with specific revision requests. WARN (partially met) -> user decides: proceed or revise
4. Integration with cook: Automatic for COMPLEX tasks, optional for STANDARD, skipped for TRIVIAL

Arguments

No CLI arguments. Takes the plan file path from conversation context (typically tasks/plans/YYMMDD-name/plan.md).

Workflow

1. Load plan file from plan directory
2. Check each of 8 dimensions against pass criteria
3. Run whole-plan consistency sweep across plan.md + all phase-*.md files (skip if plan has only plan.md)
4. Produce validation report with PASS/FAIL/WARN per dimension and specific findings
5. Route result: all PASS -> proceed; any FAIL -> return for revision; WARN -> user decides

Usage

Auto-triggered by cook workflow for COMPLEX tasks:

Gate 1 (plan approved) -> mk:validate-plan -> Phase 2 (Test)

Manual invocation:

/mk:validate-plan

Example Prompt

"I just got my plan approved. Can you validate it before we start building?"

The skill will: load the plan file, score all 8 dimensions, run the consistency sweep across phase files, and output a validation report showing pass/fail per dimension with specific findings.

Common Use Cases

• Gate check between plan approval and implementation
• Catching missing acceptance criteria before building tests
• Identifying unlisted dependencies that would block implementation
• Surfacing security oversights in user-data features
• Verifying all risks have mitigations documented

Pro Tips

• The consistency sweep catches contradictions across distributed plan files -- a phase file referencing a component that plan.md renamed, or declaring scope that plan.md marks out-of-scope
• Sweep failures do NOT auto-fail the validation -- they surface inconsistencies the human must reconcile
• This skill checks content quality, not structure. The validate-plan.py script checks plan file structure (required sections exist). Both can and should run
• Three-result output: PASS = proceed, FAIL = revise that dimension, WARN = user discretion

Notes

• The doc mentions "Offer export options for team review" which is not present in the source SKILL.md. This appears to be a doc-only addition that may need removal or source implementation.