writing-skills

Warn

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The file persuasion-principles.md provides a manual on using psychological manipulation techniques such as 'Authority' and 'Scarcity' to force agent compliance. It specifically instructs the use of imperative language like 'YOU MUST', 'Never', and 'No exceptions' to override an agent's internal reasoning and safety filters, which mirrors common prompt injection and bypass patterns.- [COMMAND_EXECUTION]: The skill includes a utility script render-graphs.js that utilizes the child_process.execSync function to run system commands (dot, which). While intended for rendering diagrams, this pattern allows for the execution of arbitrary system binaries with arguments derived from files within the skill folder.- [PROMPT_INJECTION]: The methodology described in testing-skills-with-subagents.md uses specialized triggers like 'IMPORTANT: This is a real scenario' and 'Choose and act' to create artificial pressure on subagents. These techniques are designed to test the limits of agent adherence by intentionally utilizing role-play and directive injection markers.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 6, 2026, 05:54 PM
Security Audit — agent-trust-hub — writing-skills