Validate a Feature Direction Before Building

April 8, 2026 · View on GitHub

Use this when

You have a feature idea but aren't sure if the approach is sound. You want industry best practices and deep research to sharpen the direction before writing any code.

Core skills

SkillRole
/best-practicesAudit against industry standards and patterns
/deep-researchMulti-source research (web + codebase + docs)
/feasibility-studyQuantitative comparison of approaches
/codex-brainstormAdversarial debate to stress-test the plan

Command flow

  1. /deep-research <topic> — gather state-of-the-art approaches, prior art, and ecosystem conventions
  2. /best-practices — audit the proposed approach against industry patterns; identifies gaps and anti-patterns
  3. Synthesize: combine research findings with best-practice gaps into 2-3 candidate approaches
  4. /feasibility-study — quantitative trade-off analysis across candidates
  5. /codex-brainstorm — adversarial debate on the top candidate; finds blind spots
  6. Decision: proceed with refined approach, or pivot based on findings

Why this combo works

Solo skillLimitationCombined effect
/deep-research aloneBroad information, no opinionated filter/best-practices filters research through proven patterns
/best-practices aloneChecks against standards but doesn't explore alternatives/deep-research provides the landscape of alternatives
/feasibility-study aloneCompares options but may miss industry contextResearch + practices provide well-informed candidates

Decision points

SituationChoice
Research reveals the idea already exists?Evaluate existing solutions; adapt rather than reinvent
Best practices audit finds anti-patterns?Redesign before investing in feasibility study
Brainstorm debate reaches no consensus?Prototype both approaches; use evidence to decide
Direction is validated?Proceed to /tech-spec/feature-dev

Gates

No enforced gates — this is an exploration workflow. The output is a validated direction with supporting evidence.

Expected outcome

  • Clear understanding of the problem landscape (research)
  • Approach validated against industry best practices
  • Quantitative comparison of alternatives
  • Blind spots identified via adversarial debate
  • Confident direction ready for /tech-spec