Home/Skills/Skill Kits/Inspect CI and skill supply-chain boundaries Curated workflowsInspect CI and skill supply-chain boundaries
Review workflow trust boundaries and inspect candidate agent skills as inert source material before execution.
How these skills work together
Map untrusted events and privileged operations in CI, then inventory the actions, scripts and skills involved. Read each selected source and dependency declaration without executing it. Trace a suspicious pattern to a concrete data-flow or permission issue, retaining exact file versions and coverage limits.
Validation and boundaries
Require evidence and conditions for each finding, distinguish legitimate examples from executable paths, and test only in a suitable isolated environment. A clean pattern scan does not certify runtime safety. Review the scanner itself before executing it, and keep recommendations separate from permission to change credentials or deploy.
Security86/100
getsentry/skills
An exploitation-focused GitHub Actions review for untrusted triggers, credentials and workflow dependencies.
First-party sourceApache-2.0Editorial review
Repository star change: +5 over 41.38 hours. 2026-10-31
Security86/100
getsentry/skills
A source-inspection workflow for agent skills that separates mechanical pattern matches from behavioral judgment.
First-party sourceApache-2.0Editorial review
Repository star change: +5 over 41.38 hours. 2026-10-31
Security86/100
trailofbits/skills
A dependency-risk workflow examining resolved versions, advisories, upstream maintenance, publisher concentration and install-time scripts.
First-party sourceCC-BY-SA-4.0Editorial review
Repository star change: +38 over 41.38 hours. 2026-10-31