self-commenting
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands to scan the codebase for markers. * Evidence (SKILL.md): 'grep -rn AICODE- ' used to retrieve session context. * Evidence (references/conventions.md): Various grep commands used to filter markers by type or keyword.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from code files to guide the agent's behavior. * Ingestion points: SKILL.md (Step 1: Scan) and references/conventions.md (Grep Commands section). * Boundary markers: Absent. Markers are interpreted directly as context or instructions. * Capability inventory: File system access and shell command execution. * Sanitization: Absent. No verification of marker content source or integrity.
Audit Metadata