name	project-skill-finder
description	Route work in this repository to the right project docs, architectural rules, and installed Codex skills. Use when a task is ambiguous, spans frontend and backend, touches the harness workflow, or needs the right skill and convention stack for this hackathon project.

Project Skill Finder

Use this skill when the request is broad or it is unclear which repo docs and external skills should guide the work.

Workflow

Read docs/executable-plan.md for product scope, stack, APIs, and delivery order.
Read docs/conventions.md for implementation guardrails.
Read docs/operator-checklist.md if the task touches deployment, accounts, secrets, MCP setup, or demo operations.
Map the task to the right installed skills:
- openai-docs for OpenAI and Codex docs
- frontend-skill for web UI implementation
- playwright for browser tests and demo validation
- security-best-practices for shell execution safety, secrets, and patch review
- cli-creator for CLI commands and local developer workflows
- vercel-deploy for frontend deployment
- render-deploy for backend deployment
Keep all work aligned with these repo truths:
- the product is a code-change verification harness, not a generic agent platform
- V1 break classes are logic regression, contract violation, and invariant violation
- advisory design critiques are not blocking failures in V1
- the primary demo loop is break, evaluate, repair, validate, and summarize

Decision Rules

If the task is frontend-heavy

Use the web implementation plan and design conventions first. The UI should prioritize timeline, evaluator state, diff, and merge confidence over chat.

If the task is backend-heavy

Protect determinism, evaluator clarity, provider abstraction, and artifact capture before adding complexity.

If the task is deployment-heavy

Keep the split simple:

frontend on Vercel
API on Render

If the task is product-ambiguous

Prefer features that strengthen the harness thesis:

clearer evaluators
better evidence
cleaner provider abstraction
stronger replayability

Reject features that make the product look like a generic coding chatbot.

Project Skill Finder

Use this skill when the request is broad or it is unclear which repo docs and external skills should guide the work.

Workflow

Read docs/executable-plan.md for product scope, stack, APIs, and delivery order.

Read docs/conventions.md for implementation guardrails.

Read docs/operator-checklist.md if the task touches deployment, accounts, secrets, MCP setup, or demo operations.

Map the task to the right installed skills:

openai-docs for OpenAI and Codex docs
frontend-skill for web UI implementation
playwright for browser tests and demo validation
security-best-practices for shell execution safety, secrets, and patch review
cli-creator for CLI commands and local developer workflows
vercel-deploy for frontend deployment
render-deploy for backend deployment

Keep all work aligned with these repo truths:

the product is a code-change verification harness, not a generic agent platform
V1 break classes are logic regression, contract violation, and invariant violation
advisory design critiques are not blocking failures in V1
the primary demo loop is break, evaluate, repair, validate, and summarize

Decision Rules

If the task is frontend-heavy

Use the web implementation plan and design conventions first. The UI should prioritize timeline, evaluator state, diff, and merge confidence over chat.

If the task is backend-heavy

Protect determinism, evaluator clarity, provider abstraction, and artifact capture before adding complexity.

If the task is deployment-heavy

Keep the split simple:

frontend on Vercel

API on Render

If the task is product-ambiguous

Prefer features that strengthen the harness thesis:

clearer evaluators

better evidence

cleaner provider abstraction

stronger replayability

Reject features that make the product look like a generic coding chatbot.

project-skill-finder