codex-run-loop

A skill from the Blueprint SDLC framework that guides AI agents through exploring architecture decisions, evaluating alternatives, and documenting tradeoffs before implementation begins.

An SDLC workflow encoded as agent skills covering design docs, specs, planning, implementation (with TDD), refactoring, code review, PR feedback, and browser-based UI verification. Focuses on encoding process rather than constraining agents, with verification as a non-negotiable step.

A structured debugging skill that guides Claude Code through a disciplined five-step process: reproduce the problem, form a single hypothesis, run a targeted check, fix the root cause, and verify with a test. It enforces rules like fixing root causes rather than symptoms and working one hypothesis at a time.

A skill for implementing a single task or scoped change by reading the task and relevant code first, delivering the smallest useful slice, adding tests where behavior changes, and verifying with the full test suite. It enforces one task at a time with minimal blast radius.

Evaluates test coverage and fills real gaps with high-value tests. Identifies edge cases, failure paths, state changes, and integration points, then adds only the tests that would catch real bugs without changing the code under test.

Skill