Startup Evaluation Agent Skill

Startup Evaluation

<skill_contract>

<input>Company, stage, evaluation decision, customer evidence, traction, economics, team, runway, terms, and risks.</input>

<output>An evidence-weighted health, venture-suitability, and financing assessment with one top constraint and cheapest test.</output>

<done>Every score and verdict traces to evidence, fatal risks remain visible, and the next test has owner, threshold, budget, and stop.</done>

<non_goals>Investment advice by narrative, averaging away fatal risk, or treating market size, conviction, or pitch quality as demand.</non_goals>

Evaluate the company the evidence supports, not the story it tells. Distinguish business health, venture suitability, and financing readiness; they are different decisions.

Usage Template

Provide: company, stage, startup type, evaluation decision, customer, problem, product, traction, team, economics, runway, round terms, and known risks. Use the rubric in references/evaluation-rubric.md when scoring is requested.

Workflow

<intake>

Classify stage, type (SME, innovation-driven, venture-scale, hard-tech, AI-native), lens (founder, diligence, fundraising, pivot), and evidence state. Define the decision and time horizon before calculating a score.

</intake>

<unknowns_gate>

Separate facts, assumptions, self-reported claims, and missing evidence. If the target decision or company identity is unclear, return NEEDS_INPUT. Continue with missing metrics only when the output is explicitly provisional and each gap has a probe.

</unknowns_gate>

<execute>

• Rank demand evidence from belief and interviews through behavior, payment, retention, expansion, and referral.
• Score eight dimensions using stage-adjusted weights: pain/beachhead, market/timing, value step-change, PMF/traction, business model/economics, team/governance, capital/runway, and moat/risk.
• For investor work, cross-check 5T: Team, Target Market, Tech/Product, Traction, Terms.
• For AI-native or hard-tech cases, test what remains defensible as components cheapen and identify physical, regulatory, deployment, or supply-chain bottlenecks.
• Diagnose runway and whether spend buys evidence for the next milestone.
• Name the single constraint most likely to invalidate or unlock the company.
• Specify the cheapest test, threshold, owner, budget, and stop condition.

Do not average away fatal risk. A healthy cash-flow business may still be a poor venture investment; a large market cannot rescue absent demand evidence.

</execute>

<evaluate>

Trace every score and verdict to the evidence ledger. Stress-test the conclusion against churn, paid acquisition dependence, founder conflict, financing timing, and platform dependency. Calibrate confidence to the weakest decision-critical claim.

</evaluate>

Failure Protocol

• NEEDS_INPUT: the evaluation decision, stage, or company boundary is unclear.
• INSUFFICIENT_EVIDENCE: the requested verdict depends on unavailable demand, retention, economics, or terms data.
• VERIFY_FAILED: a score lacks evidence or contradicts the ledger; rescore or mark unknown.
• BUDGET_STOP: return the provisional constraint and highest-value evidence request.

Output Contract

Return status, result (verdict, scorecard, top constraint, fatal risks, and next test), evidence (fact/assumption ledger), unknowns, and next_action with owner and threshold.

Edge Cases

• Pre-revenue company has no retention data: do not assign traction maturity; score the available behavioral test and make payment/usage the next gate.
• Bootstrapped company has strong cash flow but a small market: rate business health separately from venture-scale suitability.

Success Metrics

• Stage, type, decision lens, and evidence maturity are explicit.
• Verdict confidence follows observed behavior rather than narrative polish.
• One top constraint and one cheap falsifiable test govern the recommendation.

Quality Gates

• [ ] Facts, assumptions, self-reports, and missing evidence are separated.
• [ ] Score weights fit stage and business type.
• [ ] Fatal risks are not hidden by averages.
• [ ] Runway, milestone, test threshold, owner, and stop condition are explicit.

</skill_contract>