council
Pass
Audited by Gen Agent Trust Hub on Jul 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
uv run researchkitcommand to execute its core deliberation logic, which involves calling local scripts to interact with several AI model CLI harnesses.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, creating a potential surface for indirect prompt injection attacks against the constituent LLMs. - Ingestion points: User input is ingested via the
<question>argument and the--context-fileparameter, which reads external file content into the agent's context (SKILL.md). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands were identified in the instructions for handling the input.
- Capability inventory: The skill possesses the ability to execute shell commands using the
uvtool and communicates with external AI services (Claude Code, Codex, Antigravity, Grok CLI) (SKILL.md). - Sanitization: The documentation does not describe any sanitization, escaping, or validation steps applied to the user input before it is processed by the model council.
Audit Metadata