executing-plans

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No patterns attempting to bypass safety filters, override system instructions, or extract system prompts were found.
  • [DATA_EXFILTRATION]: No network operations, hardcoded credentials, or access to sensitive local file paths (e.g., .ssh, .aws) are present.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. It references other workflow skills within the 'superpowers' namespace for git operations and task completion.
  • [COMMAND_EXECUTION]: While the skill's purpose is to execute implementation tasks, it mandates a critical review phase (Step 1) and includes clear 'Stop and Ask for Help' triggers to prevent unintended actions or guessing.
  • [SAFE]: The skill incorporates safety best practices, such as requiring explicit user consent before modifying main/master branches and emphasizing human-in-the-loop verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 05:06 PM
Security Audit — agent-trust-hub — executing-plans