skill-guide
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected in the skill's instructions or metadata.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection as it is designed to ingest and analyze external skill manifests, catalogs, and local project paths to provide recommendations.
- Ingestion points: The skill reads skill catalogs and manifests from the environment and project-local paths (
SKILL.md). - Boundary markers: The instructions explicitly state that reading candidates is for "advisory inspection" and does not grant authorization to execute, which helps prevent the agent from accidentally following instructions inside a manifest.
- Capability inventory: The skill has no capabilities to execute code, write files, or perform network operations. It is strictly limited by the
disable-model-invocation: trueplatform configuration. - Sanitization: No specific sanitization or filtering of external manifest content is present, but the lack of execution tools limits the potential impact.
Audit Metadata