GPR-008 · medium risk

Allow a reviewed skill

Allow a specific skill identifier after its source and behavior have been reviewed.

Decision

allow

Matcher

tool = trusted-skill

Reviewed

2026-08-09

Safe test cases

matching synthetic case

tool = trusted-skillallow

different benign synthetic case

tool = benign-trusted-skillunmatched

Limitations

  • Publisher trust and skill content can change; re-review on version or hash change.

Review before applying

This recipe is a starting point, not a universal security policy. Open it in Policy Studio, replace example identifiers where necessary, review scope and blast radius, simulate where supported, and use the normal approval flow before enforcement.

Recipe SHA-256: 6c65212f9f720b8e293e523156f9dc8a43f9a9c8be8d5f94583e853ffd940a18

When a policy is saved and delivered through Guard Cloud, it uses the existing Guard policy-bundle compiler. If the policy-bundle signing key is configured, that compiler produces an RSA-PSS-SHA256 signed bundle that local Guard verifies before applying.