logos-distributed-reasoning-router

Fail

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs users to clone a code repository from an unverified third-party GitHub account (https://github.com/rak7777/mythic-mcp-proxy.git).
  • [REMOTE_CODE_EXECUTION]: The installation instructions require executing pip install -e . on the downloaded repository. This executes arbitrary code from the third-party source during the setup process.
  • [INDIRECT_PROMPT_INJECTION]: The skill demonstrates patterns for ingesting untrusted external data which could contain hidden malicious instructions.
  • Ingestion points: Data is read from external files including papers/ai_safety_paper.pdf, data/large_dataset.csv, and api.py (SKILL.md).
  • Boundary markers: The code examples do not implement any delimiters or 'ignore previous instructions' warnings when processing these files.
  • Capability inventory: The skill possesses capabilities to perform network requests (requests.post, websocket.create_connection) and write to the file system, which could be leveraged by an injected prompt.
  • Sanitization: There is no evidence of content sanitization or validation before the ingested data is passed to the AI reasoning engine.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 16, 2026, 10:44 PM
Security Audit — agent-trust-hub — logos-distributed-reasoning-router