The Agent Skills Directory

[PROMPT_INJECTION]: The file persuasion-principles.md and related sections in SKILL.md provide a framework for using psychological techniques (Authority, Commitment, Scarcity, Social Proof) to force agent compliance. These instructions are designed to override the agent's natural reasoning and safety guardrails by using imperative language like "YOU MUST", "Never", and "No exceptions" to eliminate "decision fatigue and rationalization."
[COMMAND_EXECUTION]: The utility script render-graphs.js uses child_process.execSync to execute the system command dot (Graphviz). This allows the agent to spawn external processes on the host system.
[DATA_EXPOSURE]: The script render-graphs.js performs file system operations, including reading and writing files based on directory paths passed as arguments, which could be used to access or modify data outside the immediate skill context if not properly constrained.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process and optimize other skill documentation. The "Language Density Contract" and "Skill Search Optimization" (SSO) rules create a surface where instructions could be compressed or structured in a way that prioritizes obedience over safety, potentially amplifying malicious prompts in the files being processed.

writing-skills