The Agent Skills Directory

COMMAND_EXECUTION (MEDIUM): The script render-graphs.js executes the system dot command via child_process.execSync. It processes content extracted from SKILL.md by passing it through stdin. While it avoids shell interpolation, it still facilitates the execution of a system utility on untrusted data.
PROMPT_INJECTION (MEDIUM): The files persuasion-principles.md and CLAUDE_MD_TESTING.md implement and advocate for the use of 'Authority', 'Scarcity', and 'Commitment' principles to force agent compliance. These patterns include directives like 'YOU MUST' and 'No exceptions' aimed at overriding the agent's default decision-making logic and bypassing safety protocols.
EXTERNAL_DOWNLOADS (LOW): The render-graphs.js script identifies missing system dependencies and recommends manual installation commands for the user but does not automate the download process.

writing-skills