dispute
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill's documentation uses the keyword 'GODMODE' in its header. This term is frequently used in adversarial contexts to bypass AI safety constraints and can trigger security detection systems.
- [DATA_EXFILTRATION]: The skill gathers data from the local project (using 'Read', 'Glob', and 'Grep') and transmits this information to external AI services (OpenAI via Codex and Google via Gemini) to perform its primary function of cross-model debate.
- [COMMAND_EXECUTION]: The skill automates the use of multiple file system tools and external AI model interfaces to perform a multi-round analysis and negotiation workflow.
- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection. 1. Ingestion points: Local project files are ingested using 'Read', 'Glob', and 'Grep' tools. 2. Boundary markers: The skill does not employ explicit delimiters or 'ignore' instructions to isolate the ingested file content from the instructions sent to external models. 3. Capability inventory: The agent has the ability to read and write files (including 'MEMORY.md') and communicate with multiple external LLMs. 4. Sanitization: The skill passes raw file content directly into prompts for other models without any validation or sanitization.
Audit Metadata