The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation, particularly in persuasion-principles.md and testing-skills-with-subagents.md, explicitly uses persuasion frameworks like "Authority" and "Commitment" to override the agent's internal reasoning and safety filters (referred to as "rationalization"). It employs imperative markers such as "YOU MUST", "Never", and "No exceptions" to ensure strict compliance with defined protocols.
[PROMPT_INJECTION]: The skill's primary function involves reading and acting upon external, potentially untrusted skill documentation files during the testing phase. This ingestion of untrusted data into the agent's instruction context creates an attack surface for indirect prompt injection, as there are no explicit boundary markers or sanitization procedures described for the processed content.
[COMMAND_EXECUTION]: The utility script render-graphs.js utilizes the Node.js child_process.execSync function to execute system commands. Specifically, it invokes the dot (Graphviz) utility to render SVG diagrams, passing content extracted from markdown blocks directly to the subprocess via stdin.

writing-skills