Skill Sentinel
Data & Analytics · v1.0.0
LiveLive factsWhite-label
- Purpose
- A linter and security scanner for agent skills — A–F grades across 6 finding categories with machine-readable findings for CI gates.
- Who it’s for
- Agent builders shipping skills that touch approvals and secrets
- Data truth
- Live facts — this product serves live-verified data maintained by CWI. Any sample, demo, or editorial content inside the product itself is labeled where it appears.
- White-label
- Licensable for white-label use — see the draft licensing terms.
What it does
A skill linter and security scanner for agent SKILL.md/AGENTS.md files — product #30 of the gear line: A–F grades across 6 finding categories (injected directives, secrets, approval-disabling, exfiltration, embedded secrets and more), with machine-readable findings for CI/install-time gates and an honest-limits panel on every scan.
Note: A linter, not a certificate of safety — the honest-limits panel says what a scan cannot prove. Stdlib-only, no network, no telemetry.
- Live product
- https://cumulativewebinc.github.io/cwi-skill-sentinel/
- Item card (JSON)
- https://cumulativewebinc.github.io/cwi-skill-sentinel/schema/findings.schema.json
Licensing
Request a white-label license
← Back to the shelf