Runs the idea-auditor scoring pipeline on an IDEA.md or IDEA.json file. Validates inputs, grades evidence from STATE/, calculates confidence-weighted dimension scores using scientific frameworks (JTBD, TAM, Rogers, Mayer), and produces a scorecard with a PROCEED/ITERATE/KILL decision, top blockers, and next tests. Invoke when the user wants to evaluate an idea, MVP, or evolving project against evidence.
Trigger this skill when the user asks to "score my idea", "evaluate this project", "run the idea scorecard", "assess my idea", or invokes /idea-auditor:score.
/idea-auditor:score <path> [--mode OSS_CLI|B2B_SaaS|Consumer_Viral|Infra_Fork_Standard]
<path>: directory containing IDEA.md (or IDEA.json), optional STATE/, optional BLUEPRINT.md--mode: scoring weight profile (default: OSS_CLI)Invoke the orchestrator agent, which executes:
scripts/validate_inputs.py --idea <path>/IDEA.mdscripts/grade_evidence.py --evidence <path>/STATE/ --out REPORTS/evidence-<DATE>.jsonscripts/calc_scorecard.py --idea <path>/IDEA.md --evidence REPORTS/evidence-<DATE>.json --mode <MODE> --out REPORTS/scorecard-<DATE>.jsonREPORTS/scorecard-YYYYMMDD.json — machine-readable scorecardREPORTS/evidence-YYYYMMDD.json — graded evidence with ConfDim values## Idea Scorecard — <IDEA name> (<DATE>)
Mode: OSS_CLI | Score: 63/100 | Confidence: 0.55 | Decision: ITERATE
### Dimensions
| Dimension | Score (0-5) | Confidence | Effective |
|-----------|------------|-----------|---------|
| wedge | 4.0 | 0.8 | 3.2 |
| friction | null | 0 | — | ← needs experiment
| ...
### Top 3 Blockers
1. No evidence of friction/TTFV — run activation test
2. Loop signal weak — no referral events tracked
3. Trust unvalidated — no commitment-tier evidence
### Next Tests
- Fake-door signup to measure wedge pull (target: >15% SR)
- Time-to-first-value measurement in onboarding (target: <10 min)
calc_scorecard.py.needs_experiment=true.validate_inputs.py fails, stop and fix inputs first.agents/orchestrator.md — full pipelineschemas/scorecard.schema.json — output contractreferences/rubric.md — 0–5 anchors per dimensionreferences/thresholds.yml — weights and gates by mode