clean-skill

name	clean-skill
description	A minimal, well-formed skill used as the security-scanner happy path.
allowed-tools	["Read","Grep"]
publisher	agentguard
signed	true

Clean Skill

This skill performs a read-only summary of a target file and produces a short structured report. It declares only Read and Grep (both inside the default tool allowlist), claims the agentguard first-party publisher, and ships a signed: true flag so the Phase-1 signature stub passes.

Behaviour

1. Read the target file.
2. Search it for the pattern provided in the skill input.
3. Emit a JSON object with path, pattern, match_count.

No code execution, no network egress, no PII.