skill
chaos-monkey-skill
Agent Skill: chaos engineering on a diff. Disprove a ticket's steady-state invariants after green tests.
Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/avaxML/chaos-monkey-skill itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 22, 2026chaos-monkey
An Agent Skill for AI coding agents. After a change's focused tests pass, it runs a bounded, hypothesis-driven pass that tries to disprove the change's invariants with real inputs. Every reproducible failure becomes a regression test.
It is not a style review, not a spec audit, and not production chaos engineering.
Install
npx skills add avaxML/chaos-monkey-skill
Works with Claude Code, Cursor, Codex, and other Agent Skills clients.
When to use
Ask for a "chaos monkey pass", "adversarial input experiments", or "another isolated pass, only executable counterexamples". Not for standards, spec completeness, performance tuning, SOLID, or debugging failing tests.
How it works
- Confirm the focused tests are green.
- Name the control invariant: the observable output that must keep holding.
- Run 3–8 experiments of the form
If we inject X, Y still holds, each with a named falsifier, an executed probe, and a state reset. - Classify each result: FIX, REJECTED, DEFERRED, NOT_REPRODUCED, UNPROVEN, or SURVIVED. Fixes get a regression test. The ledger carries into the next pass.
- Rerun the new tests, then the original baseline, then loop until a pass finds nothing new.
Agent instructions: SKILL.md. Injection catalog: variables.md.
Releases
Bump metadata.version in SKILL.md, add a CHANGELOG.md entry, and tag the
matching SemVer version. skills.sh lists and ranks skills from anonymous CLI
install telemetry; there is nothing to submit.