swarm-plan

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a planning orchestrator for feature decomposition. It defines clear operational workflows for research, classification, and task creation without introducing security vulnerabilities.
  • [DATA_EXFILTRATION]: No sensitive file access (e.g., SSH keys, credentials) or network exfiltration patterns were detected. The skill's research phase uses defined tools for library and codebase exploration.
  • [REMOTE_CODE_EXECUTION]: There are no instances of external script downloads, package installations, or piped shell execution from untrusted sources.
  • [PROMPT_INJECTION]: The instructions do not contain markers for overriding system guidelines, bypassing safety filters, or extracting system prompts. The skill includes a defensive constraint to 'ALWAYS validate arguments before using in commands'.
  • [COMMAND_EXECUTION]: Shell commands mentioned in the parallel exploration pattern are generic and focus on codebase search and dependency mapping using local tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 10:41 PM
Security Audit — agent-trust-hub — swarm-plan