Skip to content

Releases: smdysk/skillprobe

Release list

skillprobe v0.1.0

skillprobe v0.1.0 Latest
Latest

Choose a tag to compare

View all tags
@smdysk smdysk released this 11 Jul 14:23
v0.1.0
a23ad3b

Behavioral test harness for Claude Code skills: run a skill in an isolated fixture (throwaway workdir + HOME) via headless Claude Code and assert on what it actually does — response text, transcript tool calls, and filesystem effects. Supports trap scenarios (a seeded defect the skill must catch), so a skill that only passes happy-path prompts isn't considered tested.

Highlights

  • Nine assertion types (output / file / tool / no_writes_outside / max_turns_used) with trap support
  • skillprobe login stores a token locally — no environment variables to manage; env vars still work for CI
  • A conversational wrapper skill: drive the whole flow from inside Claude Code
  • Credentials are redacted from reports; each run is isolated (not a security sandbox — see the README)
  • 112-test suite; CI on Linux/macOS/Windows × Node 20/22

See the README for the quick start and the isolation model.

Assets 2
Loading