ultraqa_skill

Overview

This skill provides an autonomous QA cycling workflow that runs test, verify, fix, and repeat cycles until a defined quality goal is met. It coordinates tester, architect, and executor agents to diagnose failures and apply precise fixes across up to five cycles. The skill tracks state, outputs clear cycle progress, and enforces early-exit and cleanup rules to avoid stale sessions.

How this skill works

The skill parses a goal (tests, build, lint, typecheck, or a custom pattern) and runs the appropriate verification command. On failure it spawns an architect agent to diagnose the root cause and an executor agent to apply recommended fixes, then repeats verification. It records failures, detects repeated failure patterns, enforces max cycles and early exits, and deletes the session state file on completion or cancellation.

When to use it

• Automate CI-style recovery when tests or builds intermittently fail.
• Iterate automatically on lint or typecheck regressions until resolved.
• Run repeated verification+fix cycles for complex regressions that need diagnosis and targeted code edits.
• Apply a custom success pattern for nonstandard checks or service-level testing.
• Use as a developer productivity tool to reduce manual triage of frequent failures.

Best practices

• Define a clear, executable goal (use --tests, --build, --lint, --typecheck, or --custom with a pattern).
• Limit side effects: ensure test/build commands are deterministic and reproducible in CI-like environments.
• Allow architect recommendations to be specific and minimal — prefer targeted edits over broad refactors.
• Enable parallel diagnosis and fix preparation to reduce cycle time when possible.
• Ensure cleanup policies and state deletion are accepted by your workspace scripts to avoid stale state files.

Example use cases

• Run ultraqa --tests to iterate until all test suites pass after a complex merge.
• Use ultraqa --build in a release pipeline to auto-fix and reattempt build failures caused by minor config issues.
• Invoke ultraqa --lint to apply quick style fixes and repeat until no lint errors remain.
• Start ultraqa --typecheck to resolve TypeScript errors with architect-proposed fixes across multiple files.
• Run ultraqa --custom "DEPLOY_READY" to verify a deployable artifact contains a custom success marker.