hermes-labyrinth-observability
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to download source code from an unverified external repository (github.com/stainlu/hermes-labyrinth) which is not associated with the skill author (reason-machines).
- [COMMAND_EXECUTION]: The installation guide suggests executing several shell commands, including 'git clone' to fetch remote content and 'mkdir'/'cp' for local environment setup.
- [REMOTE_CODE_EXECUTION]: The setup process requires the user to run build and test scripts ('npm run build', 'npm test') on the downloaded code, leading to the execution of unverified remote scripts on the local system.
- [INDIRECT_PROMPT_INJECTION]: The skill processes agent logs including tool calls and prompts, which are then displayed in a dashboard. This creates a vulnerability surface where malicious instructions embedded in the logs could potentially be executed if the agent or user acts upon the visualized data. The skill notes that secret redaction is performed, which is a positive security measure, but the ingestion point for untrusted data remains.
Audit Metadata