Spec-driven agent harness for Claude Code and Codex — five-agent pipeline (scope, plan, build, verify, learn) with an auditable proof chain.
Skill Claude CodeCodex
Invoke when implementing features, writing code, or reviewing code quality. Contains project-specific naming conventions, error handling patterns, import style, and deviations from standard practices.
Skill Claude CodeCodex
Invoke when working on deployment configuration, CI/CD pipelines, environment variables, or release processes. Contains project-specific deploy platform conventions.
Skill Claude CodeCodex
Invoke before any git operations — branching, committing, merging, or creating pull requests. Contains project-specific branch naming, commit format, and merge strategy.
Skill Claude CodeCodex
Invoke when writing tests, reviewing test quality, or setting up test infrastructure. Contains project-specific testing framework conventions, fixture patterns, and coverage expectations.
Skill Claude CodeCodex
Invoke when debugging failures, diagnosing unexpected behavior, or investigating test failures. Contains project-specific failure modes, diagnostic workflows, and known issues.
Skill Claude CodeCodex
Invoke when building features that call LLM APIs, handling AI responses, managing prompts, or integrating AI SDKs. Contains error handling, security, prompt management, and observability patterns.
Skill Claude CodeCodex
Invoke when implementing API routes, request handling, middleware, or error responses. Contains validation, error format, route architecture, and authorization patterns.
Skill Claude CodeCodex
Invoke when working with database queries, schema changes, migrations, or data models. Contains project-specific ORM conventions and data access patterns.