recipe-quality-profile

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill interpolates user-supplied input ($ARGUMENTS) directly into its primary prompt instructions without using boundary markers or sanitization. It also instructs the agent to read various repository files, which could contain instructions designed to influence the agent's logic during the policy generation process.
  • Ingestion points: The $ARGUMENTS variable in SKILL.md and repository content such as docs/project-context/quality.yaml, CI manifests, scripts, and implementation patterns.
  • Boundary markers: The skill lacks explicit delimiters (like XML tags or markdown blocks with 'ignore' instructions) around the interpolated user input or the data read from the repository.
  • Capability inventory: The skill utilizes the agent's ability to read multiple file types across the file system and perform write operations to docs/project-context/quality.yaml.
  • Sanitization: No programmatic escaping or filtering is mentioned, but the authoring flow includes a mandatory human-in-the-loop confirmation step before the agent is permitted to write the resulting file to the repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 06:16 AM
Security Audit — agent-trust-hub — recipe-quality-profile