slate-ar-recipe
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local CLI tools using Node.js scripts located at
../codex-autoresearch/plugins/codex-autoresearch/scripts/autoresearch.mjs. These commands are used to list, recommend, and display setup plans. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. Evidence: 1. Ingestion points: external recipe catalogs (via
recipes list). 2. Boundary markers: instruction to inspect catalogs before trust. 3. Capability inventory: shell command execution via Node.js scripts in SKILL.md. 4. Sanitization: none mentioned. The instruction 'Do not trust external recipe catalogs without inspecting them first' identifies this risk factor.
Audit Metadata