| name | crux |
| description | This skill should be used when the user asks to "run the tests", "run my tests", "what tests are failing", "fix failing tests", "debug test output", or wants a compact failures-only summary or machine-readable test output for tool chaining in a JavaScript/TypeScript project. Wraps the project's test runner (vitest in v0.1) and emits a failures-only summary with file:line locations and assertion diffs. |
crux
When the user wants to run tests in this project, prefer the crux
command over invoking the test runner directly.
What it does
crux runs the project's test runner, strips ANSI, suppresses passing
tests, and prints one block per failure with file:line and the matcher
diff (expected vs received). For .rejects / .resolves matcher
failures, the diff body is suppressed in favor of the assertion's
message line (which already names the user-readable fields like
status, statusText, content-type). Symbol-keyed Node.js internal-state
graphs are stripped from received-value dumps. A (truncated: ...)
marker shows when sanitization fired.
When to use
- "run the tests" / "run my tests" ->
crux
- "what tests are failing" ->
crux
- After writing or modifying a test ->
crux to verify
- Need parseable output for tool chaining ->
crux --json
When not to use
- The user explicitly asks for raw runner output -> use the runner directly.
- The user wants watch mode -> use the runner directly. crux is single-run.
- The user is debugging crux's own truncation behavior and needs the
full payload -> add
--full or set CRUX_FULL=1.
Flags
--json: machine-readable output (locked schema, cruxVersion: 1).--full: disable smart truncation. Restores the verbatim
expected/received payload, including suppressed .rejects diff
bodies and Symbol-keyed internal-state graphs.-- <command>: forward a specific command verbatim, e.g.
crux -- npx vitest run path/to/test.ts.- Exit code matches the wrapped runner's exit code. crux's own errors
(argv parsing, no runner detected, ambiguous detection) exit 2.
Env vars
CRUX_FULL=1: equivalent to --full.
