writing-skills

Warn

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The script render-graphs.js utilizes execSync to invoke the Graphviz dot utility. It extracts content from SKILL.md and passes it directly to the process stdin. While intended for diagram rendering, the use of execSync on content derived from data files constitutes a risk of command injection or unauthorized execution if the input files are maliciously modified.\n- PROMPT_INJECTION (LOW): The files persuasion-principles.md and examples/CLAUDE_MD_TESTING.md contain instructions and emphatic prompt templates (e.g., Variant C) designed to override an agent's reasoning. These patterns use 'Authority' and 'Commitment' techniques (e.g., 'THIS IS EXTREMELY IMPORTANT', 'you failed') to force compliance and bypass standard agent decision-making filters.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 18, 2026, 07:28 AM