refine
Pass
Audited by Gen Agent Trust Hub on Mar 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
git diff main...HEAD --stat. This is a read-only local command used to determine which files were modified during the session.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the repository's documentation and skill files to generate updates.\n - Ingestion points:
SKILL.md(readsdocs/validation.md,docs/enterprise.md,docs/autonomous-mode.md, and other invoked skills).\n - Boundary markers: No specific delimiters or instructions are provided to the agent to distinguish between documentation content and potential instructions.\n
- Capability inventory: Executing
git diff(read-only) and writing/editing files in thedocs/directory.\n - Sanitization: No sanitization or validation of the ingested content is performed before processing.
Audit Metadata