Runs adversarial end-to-end QA cycles with hostile scenario matrices and structured failure reports.
npx skills add https://github.com/catalystctl/catcode --skill ultraqa-catalystctl
Please help me install this Agent Skill. Skill: ultraqa Source: https://github.com/catalystctl/catcode/tree/main/.omp/skills/ultraqa Command: npx skills add https://github.com/catalystctl/catcode --skill ultraqa-catalystctl
High-intent search queries and answers about installing and using this skill.
Invoke the skill with a goal flag such as --tests, --build, or --custom. It builds a scenario matrix covering malformed input, injection, and stale state, then runs baseline commands plus adversarial harnesses and reports failures with evidence.
Unit tests verify isolated logic under expected conditions, while adversarial e2e testing models hostile users, interruptions, and corrupted state against the running system. This skill treats green unit tests as insufficient and requires behavior-level scenario evidence.
The scenario matrix includes prompt injection rows covering override instructions, secret exfiltration attempts, and false success claims. Each row runs a harness with an expected signal, and results are recorded with evidence in the report.
Yes. Flaky test rows use rerun strategies and failure clustering to avoid false greens from one lucky pass. Hung commands are bounded by timeouts with killed-child cleanup and recovery notes.
The loop stops when the goal is met, after 5 cycles, when the same failure signature appears 3 times, or when a safety bound blocks progress. Each stop condition produces a structured exit message with residual risks.
LudwigAJImplement Rust features, refactors, and bug fixes with test-first validation.
happier-devOrchestrates evidence-backed code, plan, and release reviews with QA coverage ledgers and finding triage.
gelsonsimoesRun linting and type checks on Node.js, TypeScript, and Python projects.
rainoftimeGenerate random inputs and shrink counterexamples to verify software invariants.
