By name: pm-verification description: "Internal skill. Use cc-p4p-router for all PM tasks." allowed-tools: Read, Grep, Glob
PM Verification
Overview
Claiming PM work is complete without verification is dishonesty, not efficiency.
Core principle: Evidence of completeness before claims, always.
The Iron Law
NO COMPLETION CLAIMS WITHOUT EVIDENCE OF COMPLETENESS
If you haven't checked the deliverable against criteria, you cannot claim it's done.
The Gate Function
BEFORE claiming any status or expressing satisfaction:
1. IDENTIFY: What criteria prove this claim?
2. CHECK: Verify each criterion against the deliverable
3. SCORE: sections_present / sections_required
4. VERIFY: Does the score justify the claim?
- If NO: State actual status with gaps
- If YES: State claim WITH evidence
5. ONLY THEN: Make the claim
Skip any step = incomplete work
Per-Workflow Checklists
SPEC Verification
| Check | Criteria | Required? |
|---|---|---|
| Problem statement | Present, 2-3 sentences, evidence-grounded | YES |
| Goals | 3-5 measurable outcomes listed | YES |
| Non-goals | 3-5 explicit exclusions with rationale | YES |
| User stories | Standard format with specific user types | YES |
| P0 requirements | Must-haves clearly identified | YES |
| Success metrics | Leading + lagging with targets | YES |
| Open questions | Tagged with owner (eng/design/legal/data) | YES |
| Acceptance criteria | Given/When/Then or checklist format | RECOMMENDED |
| Timeline | Hard deadlines and phasing noted | OPTIONAL |
Completeness score: Count YES criteria present / total YES criteria required. Pass threshold: 6/7 (all YES criteria) = SPEC_COMPLETE. Missing any = NEEDS_CLARIFICATION.
RESEARCH Verification
| Check | Criteria | Required? |
|---|---|---|
| Sources cited | Each finding has named sources | YES |
| Evidence graded | Strong (3+) / Medium (1-2) / Weak (assumption) for each theme | YES |
| Themes triangulated | Multiple data points per theme | YES |
| Opportunities sized | Users affected x frequency x severity | YES |
| Methodology stated | How data was collected and analyzed | YES |
| Contradictions noted | Where sources disagree, documented | RECOMMENDED |
| Personas identified | If data supports, behavioral clusters described | OPTIONAL |
Pass threshold: 5/5 = SYNTHESIS_COMPLETE. Missing any = NEEDS_MORE_DATA.
ROADMAP Verification
| Check | Criteria | Required? |
|---|---|---|
| Items scored | Each item scored with chosen framework | YES |
| Dependencies mapped | Owner + need-by date for each dependency | YES |
| Capacity considered | Items fit within team capacity constraints | YES |
| Strategy-aligned | Each item serves a stated goal/OKR | YES |
| Now/Next/Later clear | Clear differentiation between time horizons | YES |
| Shipped acknowledged | Recent completions documented | RECOMMENDED |
Pass threshold: 5/5 = ROADMAP_UPDATED. Missing any = NEEDS_CLARIFICATION.
COMMUNICATE Verification
| Check | Criteria | Required? |
|---|---|---|
| Audience matched | Tone, detail level, format match audience | YES |
| No jargon leaks | Internal terms not in customer comms | YES |
| Actionable | Clear next steps or asks | YES |
| Status justified | Green/Yellow/Red backed by evidence | YES (if status update) |
| Length appropriate | Under limits (exec: 200 words, eng: flexible) | YES |
Pass threshold: 4/4 (or 5/5 for status updates) = DRAFT_READY.
Evidence Grounding
For every claim in a PM deliverable, ask: "What data supports this?"
Strong Evidence
- User research data (interviews, surveys, analytics)
- Competitive analysis with feature comparison
- Metrics with statistical significance
- Multiple independent sources agreeing
Medium Evidence
- Expert opinion from domain SMEs
- Single data source (one survey, one analysis)
- Customer feedback from small sample
- Industry benchmarks
Weak Evidence (Flag These)
- Assumptions without data
- "Common sense" claims
- Extrapolation from unrelated contexts
- Single anecdote
Rule: Every finding should state its evidence strength. Weak evidence must be flagged explicitly.
Red Flags — STOP
If you find yourself using:
- "Seems fine" — What specifically is fine? Check the list.
- "Probably" — Based on what evidence?
- "Should work" — For whom? Verified how?
- "Users will love it" — What data supports this?
- "This is obvious" — Obvious to whom? Document it anyway.
STOP. Check against the appropriate workflow checklist above.
Rationalization Prevention
| Excuse | Reality |
|---|---|
| "It's good enough" | Check the completeness score. Is it really? |
| "We can refine later" | Ship complete, refine content. Don't ship incomplete structure. |
| "The PM will know" | Document explicitly. Context gets lost. |
| "Minor gaps don't matter" | Minor gaps become major confusion in implementation. |
| "I'm confident" | Confidence does not equal completeness. Check the list. |
Completeness Scoring
COMPLETENESS = sections_present / sections_required
SPEC: [X/7] — Problem, Goals, Non-Goals, Stories, P0 Reqs, Metrics, Questions
RESEARCH: [X/5] — Sources, Grading, Triangulation, Sizing, Methodology
ROADMAP: [X/5] — Scored, Dependencies, Capacity, Strategy, Horizons
COMMS: [X/4+] — Audience, Jargon, Actionable, Length (+Status if applicable)
Self-Critique Gate (BEFORE Router Contract)
MANDATORY: Check these BEFORE writing Router Contract:
Deliverable Quality
- Meets all required criteria for this workflow type?
- Evidence strength stated for key claims?
- No placeholder text or TODOs remaining?
- Saved to correct output location?
Self-Critique Verdict
PROCEED: [YES/NO] CONFIDENCE: [High/Medium/Low] COMPLETENESS: [X/Y]
- If NO — Fix gaps before writing Router Contract
- If YES — Proceed to Router Contract
Output Format
## Verification Summary
### Workflow
[SPEC | RESEARCH | ROADMAP | COMMUNICATE]
### Criteria Checked
| Check | Status | Evidence |
|-------|--------|----------|
| [Criterion] | PASS/FAIL | [What was verified] |
### Completeness
[X/Y] criteria met
### Gaps (if any)
- [Missing element] — [Severity: Blocking/Non-blocking]
### Verdict
COMPLETE — All required criteria verified