The Agent Skills Directory

[COMMAND_EXECUTION]: The render-graphs.js utility script uses child_process.execSync to execute system commands. It invokes the dot binary to render diagrams from markdown files and uses which to verify the command's availability. While intended for visualization, executing subprocesses based on file content poses a security risk if the input is not strictly controlled.- [PROMPT_INJECTION]: The skill teaches methodologies for enforcing rigid agent behavior that mirror prompt injection techniques. The persuasion-principles.md and SKILL.md files guide authors to use "Authority" and "Commitment" principles (e.g., using terms like "YOU MUST" and "No exceptions") to override an agent's standard reasoning and autonomy. These patterns are designed to limit the agent's discretion and ensure compliance, which is a technique frequently exploited in behavioral-bypass attacks.- [INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface by processing external skill files. Ingestion points: The render-graphs.js script reads SKILL.md files from user-provided directories. Boundary markers: None are implemented; the script relies on standard markdown code fences to extract content. Capability inventory: The script possesses the capability to execute shell commands (dot). Sanitization: No sanitization is performed on the extracted Graphviz code before it is passed to the system shell, potentially allowing for the exploitation of vulnerabilities in the rendering engine.

writing-skills