skills/lev-os/agents/lev-find/Gen Agent Trust Hub

lev-find

Warn

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes multiple shell commands to perform searches, including find, grep, jq, and curl. It also executes a local JavaScript file located at ~/lev/workshop/poc/lookup/cli.js.
  • [DATA_EXFILTRATION]: The skill accesses sensitive local file paths, specifically session history logs at ~/.clawdbot/agents/main/sessions. It also mandates sending user queries and research context to external, non-whitelisted third-party APIs (Exa and Valyu). This combination creates a risk where sensitive information stored in session history or user queries could be sent to external research backends.
  • [EXTERNAL_DOWNLOADS]: The skill requires integration with external third-party services (Exa, Valyu, Brave Search, etc.) that are not part of the trusted vendors list. It explicitly mandates loading these backends for any research-scoped query.
  • [PROMPT_INJECTION]: The skill includes a 'Prompt Architect Overlay' section that uses role-play techniques to define a specific operating persona ('prompt-architect-enhanced specialist') and set of 'contracts' for the agent, which could be used to influence or override standard agent behavior logic.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 1, 2026, 09:43 AM