grimoire-loop

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as tickets, specifications, and code diffs, which can contain hidden instructions that might influence the agent's logic.
  • Ingestion points: User-provided specifications and repository diffs mentioned in SKILL.md and references/agent-prompts.md.
  • Boundary markers: The prompt templates in references/agent-prompts.md do not utilize strict delimiters or safety instructions to prevent the agent from following directions within untrusted data.
  • Capability inventory: The skill possesses file-writing capabilities and can execute local formatting commands.
  • Sanitization: No evidence was found of input sanitization or validation methods to mitigate injection risks.
  • [COMMAND_EXECUTION]: The skill facilitates the use of local command-line tools for automated code formatting.
  • Evidence: references/format-detection.md defines command strings for various tools like prettier, ruff, and black.
  • Mitigation: The instructions require explicit authorization for writes and advise using the project's pinned tool versions.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 11:07 AM