superpowers-writing-skills

Warn

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill documentation in testing-skills-with-subagents.md and persuasion-principles.md provides templates for influencing agent behavior using authoritative overrides. It explicitly recommends patterns such as IMPORTANT: This is a real scenario. Choose and act. and absolute imperatives like YOU MUST and No exceptions. These techniques are designed to bypass an agent's standard reasoning process and safety guidelines to ensure compliance under simulated pressure. \n- [COMMAND_EXECUTION]: The render-graphs.js utility script uses child_process.execSync to invoke the system dot command (Graphviz). The script extracts content from SKILL.md and pipes it as standard input to the system process. Executing system commands with content derived from user-provided files constitutes a security risk, particularly in environments where input sanitization is not strictly enforced.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 15, 2026, 08:18 AM