The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill contains explicit strategies and examples for manipulating agent behavior using 'Authority' and 'Commitment' principles.\n
Evidence in persuasion-principles.md: Guides users to use imperative language like "YOU MUST", "Never", "Always", and "No exceptions" to eliminate "decision fatigue and rationalization".\n
Evidence in examples/CLAUDE_MD_TESTING.md: Scenario variants use high-pressure language ("If a skill existed... and you didn't use it, you failed") and markers like "IMPORTANT: This is a real scenario. Choose and act." designed to override the agent's normal decision-making process.\n- [COMMAND_EXECUTION] (MEDIUM): The render-graphs.js script executes the system dot command using execSync on content extracted from markdown files.\n
Evidence: execSync('dot -Tsvg', { input: dotContent, ... }) in render-graphs.js line 67.\n
Risk: If the installed Graphviz version is vulnerable or configured unsafely, malicious Graphviz code embedded in a SKILL.md file could lead to unauthorized system access or file reads.

writing-skills