subagent-context-retrieval

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill provides instructional guidelines for agents to evaluate file relevance and refine search queries. These instructions are functional and do not attempt to override safety filters or system constraints.
  • [DATA_EXPOSURE]: The skill discusses searching for files related to authentication and sessions as part of its technical examples. It does not contain hardcoded credentials, access sensitive system paths (e.g., ~/.ssh), or perform unauthorized data access.
  • [DATA_EXFILTRATION]: There are no network operations, curl/wget commands, or data transmission logic present in the skill content.
  • [REMOTE_CODE_EXECUTION]: No external dependencies, script downloads, or dynamic execution patterns are used. The JavaScript snippets provided are conceptual pseudocode examples for illustrative purposes.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 10:28 AM
Security Audit — agent-trust-hub — subagent-context-retrieval