logos-router-distributed-reasoning
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install Ollama, a well-known service for local inference, using its official installation script from
https://ollama.com/install.sh. - [REMOTE_CODE_EXECUTION]: The installation process for the prerequisite inference engine uses a pipe-to-shell pattern (
curl -fsSL ... | sh). While this allows for remote code execution, it is the standard installation method for a recognized and well-known service. - [EXTERNAL_DOWNLOADS]: The skill clones a third-party repository from GitHub (
https://github.com/rak7777/mythic-mcp-proxy.git) to install the core Logos Router functionality. This is a standard part of the skill's deployment process. - [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it processes untrusted user data and external files (e.g., research papers and source code) to perform reasoning tasks.
- Ingestion points: Untrusted data enters the system through
router.query()prompts andworkflow.execute()inputs defined inSKILL.md. - Boundary markers: No explicit boundary markers or 'ignore' instructions are documented in the usage examples.
- Capability inventory: The skill interacts with local inference APIs (Ollama, VLLM) and performs file system read operations for document analysis.
- Sanitization: No input sanitization or validation logic is visible in the provided code snippets.
Audit Metadata