trim-cot-leakage
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill involves reading and modifying repository files (documentation, code comments, and notes) based on their content. This ingestion of potentially untrusted data creates a surface for indirect prompt injection. However, the risk is mitigated by a rigorous workflow that requires the agent to judge every match semantically, enumerate all logical propositions before editing, and follow specific 'keep rules' to prevent overcorrection.\n- [COMMAND_EXECUTION]: The skill uses the
ripgrep(rg) command-line tool to identify suspect patterns within the repository. These searches are restricted to project files and explicitly exclude sensitive system directories like.git/, ensuring the execution is scoped to the intended audit area.
Audit Metadata