garfield
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted policy documentation.
- Ingestion points: Discovers and reads
policies/**/*.mdfiles from the target repository at runtime (documented inSKILL.md). - Boundary markers: Prompt templates for subagents lack clear delimiters or instructions to ignore embedded commands within the injected policy text.
- Capability inventory: The coordinator agent can execute file system writes to implement accepted fixes based on subagent output.
- Sanitization: No validation or sanitization is performed on the ingested markdown content before it is processed.
Audit Metadata