startup-validator（启动验证器）

@rules/evidence-and-scoring.md @references/frameworks.md

Startup Validator

Score startup ideas with explicit evidence and uncertainty instead of optimism. Organize results as a multi-file folder for future reference.

<when_to_use>

Use this skill when:

• validating a new startup or product idea
• deciding whether to proceed, pivot, or narrow the market
• preparing for customer discovery or fundraising conversations

Do not use this skill when:

• the main job is generating many new ideas
• the request is technical implementation planning
• the user wants first-principles redesign rather than validation scoring

Boundary:

• use genius-thinking for broad ideation
• use elon-musk for assumption teardown and redesign

Examples:

/startup-validator AI-based education service
/startup-validator subscription healthcare app
/startup-validator B2B purchasing automation

</when_to_use>

<input_check>

If $ARGUMENTS is missing, ask:

Which startup idea should we validate?

If founder, market, or customer evidence is missing, continue with explicit assumptions instead of inventing certainty.

</input_check>

<owned_job>

For each run:

1. Restate the idea and identify the key unknowns.
2. Score the idea using the framework set in references/frameworks.md.
3. Mark evidence quality and uncertainty for each major claim.
4. Summarize the strongest strengths, critical weaknesses, and next validation steps.

</owned_job>

<document_shape>

Output Structure

.hypercore/startup-validator/[topic-slug]/
├── flow.json           # phase tracking
├── thesis.md           # idea framing + key hypotheses
├── thiel-scores.md     # per-question rationale and scores
├── pmf-forces.md       # PMF checklist + Forces of Progress
└── verdict.md          # final score + grade + weaknesses + roadmap

• Use ASCII kebab-case for [topic-slug] (e.g., ai-education-service).
• Each phase produces its own file for organized reference.
• flow.json tracks progress through phases. See references/flow-schema.md for the schema.
• If the folder exists from a prior run, read existing files before updating.

</document_shape>

<flow_tracking>

Flow Tracking

Write flow.json at the start and update as each phase completes.

Phase progression

Resume support

If flow.json already exists, read it and continue from the last incomplete phase. Do not restart completed phases.

</flow_tracking>

Workflow

Scoring rule:

• do not fabricate precise scores when evidence is thin
• mark uncertain sections as provisional and explain what evidence would change them

<output_contract>

Each output file should follow the formats defined in the KO version's result_structure section:

• thesis.md: one-line thesis, value hypothesis, growth hypothesis, key unknowns
• thiel-scores.md: per-question score + rationale + improvement direction
• pmf-forces.md: PMF checklist, Forces of Progress (Push/Pull/Habit/Anxiety), switching probability
• verdict.md: total score/grade, critical weaknesses with severity, improvement roadmap (immediate/30d/90d)

</output_contract>

Before finishing, check:

• unsupported certainty is called out as provisional
• evidence quality is distinguished from opinion
• the score ties back to named frameworks
• the output includes concrete next validation actions
• all output files are saved under .hypercore/startup-validator/[topic-slug]/
• flow.json status is set to completed