Function runHook

vigiles/testing — Pillar 2 entry point: the harness-testing API. Re-exports the three tiers — runHook (unit), runHarnessTest (deterministic), runEval (eval) — plus the runner-agnostic predicates/assertions. Kept deliberately separate from vigiles/claude-code so this surface can stay harness-agnostic as more harnesses are added. Granular paths (vigiles/run-hook, etc.) still work.

It re-exports the composition-root runner modules (which do the Claude-Code default-wiring), never an adapter directly — the agnostic-surface eslint boundary forbids importing src/adapters/* from here. See research/adapter-api-design.md.

runHook(command: string, input: HookInput, opts?: RunHookOptions): HookRunResult
Run a hook command, piping input as JSON to its stdin, and report the exit code + parsed decision. Synchronous (so it can be used inside an eval's measure too). command is run through a shell, so the same command string a plugin ships (with args / env refs) works verbatim. Mark a hook you didn't write with trusted: false and it is confined by default (or pass sandbox: "auto" directly) — see RunHookOptions.trusted.
Parameters
- command: string
- input: HookInput
- opts: RunHookOptions = {}
Returns HookRunResult
- Defined in src/run-hook.ts:753

Function runHook

Parameters

Returns HookRunResult

Settings