link-drop-pipeline
Fail
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill documentation and setup steps explicitly reference reading sensitive bearer tokens and API keys from local configuration files located at
$HOME/Mundi Princeps/config/api_keys.json. - [COMMAND_EXECUTION]: The skill relies heavily on executing numerous shell scripts (
route.sh,analyze-youtube.sh,build-registry.sh) and system utilities (gh repo view,git commit,rsync) with arguments derived from external, untrusted URL content. - [DATA_EXFILTRATION]: The skill uses
rsyncandsshto synchronize the contents of the local~/.claude/skills/directory to a remote server (achilles-mundi). While this is part of its stated 'Achilles' sync workflow, it creates a mechanism for sensitive data to leave the local environment. - [PROMPT_INJECTION]: The skill is highly vulnerable to Indirect Prompt Injection (Category 8) because it fetches and processes arbitrary text from YouTube, X (Twitter), GitHub, and blog posts, then uses an LLM (Gemini) to extract insights and generate new skill files without explicit boundary markers or sanitization of the fetched content.
Recommendations
- AI detected serious security threats
Audit Metadata