Agent skill

recipe-validate

Orchestrate hypothesis validation through type-appropriate methods — prototypes, code analysis, market research, and expert review

Stars 1
Forks 1

Install this agent skill to your Project

npx add-skill https://github.com/shinpr/claude-code-discover/tree/main/skills/recipe-validate

SKILL.md

Context: Validate hypotheses using appropriate methods based on risk type. Invoke hypothesis-verifier for bias-free validation design. Record results in hypothesis files.

Orchestrator Definition

Execution Protocol:

  1. Delegate verification design to hypothesis-verifier (via Agent tool, subagent_type: "discover:hypothesis-verifier") for bias-free assessment
  2. Follow the validation flow defined below
  3. Stop at every [STOP — BLOCKING] marker — present findings and CANNOT proceed until user explicitly confirms

Workflow Overview

Input (hypothesis file or hypothesis ID)
    ↓
1. Hypothesis Assessment → Read and understand the hypothesis
    ↓
2. Validation Design → hypothesis-verifier designs the test [Stop: User confirms design]
    ↓
3. Validation Execution → Type-appropriate method
    ↓
4. Result Recording → Update hypothesis file [Stop: User reviews results]
    ↓
Output: Updated hypothesis file with validation results

Execution Decision Flow

1. Hypothesis Assessment

Input: $ARGUMENTS

Read the target hypothesis file(s). Understand:

  • What is being tested (the hypothesis statement)
  • Which risk dimension is primary (Value / Usability / Feasibility / Viability)
  • Current confidence scores
  • Time budget and deadline
  • Parent Opportunity context

2. Validation Design

Invoke hypothesis-verifier using Agent tool (subagent_type: "discover:hypothesis-verifier") to design the validation:

  • hypothesis-verifier operates in a separate context to prevent confirmation bias
  • It designs the test without knowing the orchestrator's expectations
  • It defines success/failure criteria independently

[STOP — BLOCKING] Present validation design to user for confirmation:

  • Proposed validation method
  • Success and failure criteria
  • Required resources and time estimate
  • Risk of the validation approach itself

CANNOT proceed to validation execution until user explicitly confirms the design.

3. Validation Execution

Execute validation based on the risk type and method:

Risk Type Validation Methods Tools
Value Market research, user interviews, competitive analysis, landing page test WebSearch tool, survey analysis
Usability Prototype testing, usability study, interaction analysis prototype-generator agent (context-separated)
Feasibility Code spike, architecture review, dependency analysis codebase-analyzer, worktree code spike
Viability Business model analysis, ROI calculation, regulatory review WebSearch tool, BMC/VPC analysis

Prototype Generation (for Usability validation)

Invoke prototype-generator using Agent tool (subagent_type: "discover:prototype-generator"):

  1. Pass the hypothesis file path and output path in the prompt
  2. prototype-generator operates in a separate context — it reads design context files independently and produces a natural product UI
  3. When multiple patterns need validation, invoke prototype-generator separately for each pattern
  4. Output: docs/discovery/prototypes/hypo-{id}-prototype.html (or hypo-{id}-{pattern}-prototype.html for multiple)
  5. User opens in browser for verification

Worktree Code Spike (for Feasibility validation)

When validating Feasibility through code:

  1. Invoke codebase-analyzer (via Agent tool, subagent_type: "discover:codebase-analyzer") for current architecture understanding
  2. Use Agent tool with isolation: "worktree" to run a code spike in an isolated environment
  3. The spike agent implements a minimal proof-of-concept
  4. Results (feasibility assessment, complexity, blockers) are reported back
  5. The worktree is automatically cleaned up — no code persists in the main branch

Market Research (for Value/Viability validation)

Use WebSearch tool to gather market data. See product-principles skill references/mvp-definition.md for scope assessment.

4. Result Recording

After validation execution:

  1. Update the hypothesis file:
    • Change status (validated / invalidated / inconclusive / timeout)
    • Update confidence scores with evidence
    • Record evidence and artifacts
    • Document learnings
  2. Link to validation artifacts (prototypes, data, interview notes)

[STOP — BLOCKING] Present results to user for review:

  • Validation outcome (validated / invalidated / inconclusive)
  • Updated confidence scores with evidence
  • Key learnings
  • Recommended next actions

CANNOT proceed to next hypothesis or close the workflow until user explicitly reviews.

5. Timeout Handling

When a hypothesis reaches its deadline without conclusion:

  1. Set status to timeout
  2. Present options to user:
    • Extend: Add more time with justification
    • Pivot: Modify the hypothesis based on partial evidence
    • Abandon: Stop validation, record learnings from partial evidence

Sub-agent Usage

Agent When Why (context separation benefit)
hypothesis-verifier (subagent_type: "discover:hypothesis-verifier") Always (validation design) Eliminates confirmation bias in test design
prototype-generator (subagent_type: "discover:prototype-generator") Usability validation Produces natural product UI without test infrastructure contamination
codebase-analyzer (subagent_type: "discover:codebase-analyzer") Feasibility validation Objective code analysis without hypothesis bias

Scope Boundaries

Included: Hypothesis validation design, validation execution, result recording Not included: Hypothesis generation, PRD creation, reflection/distillation

Completion Criteria

  • Hypothesis assessed and understood
  • hypothesis-verifier designed the validation (bias-free)
  • User confirmed validation design
  • Validation executed with appropriate method
  • Hypothesis file updated with results and evidence
  • User reviewed results
  • Next actions identified

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results