The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates and advocates for the use of psychological persuasion principles (Authority, Commitment, Scarcity) in documentation to enforce model compliance. These patterns include the use of imperative language (e.g., 'YOU MUST', 'No exceptions', 'The Iron Law') designed to override model rationalization and ensure adherence to critical processes.
[COMMAND_EXECUTION]: The 'render-graphs.js' utility script executes the system 'dot' binary (Graphviz) using 'child_process.execSync'. This is a legitimate function for converting markdown code blocks into SVG diagrams. The script includes basic input sanitization by matching graph names against the '\w+' regex and passing graph content via standard input.
[PROMPT_INJECTION]: The 'testing-skills-with-subagents.md' and 'examples/CLAUDE_MD_TESTING.md' files define 'Pressure Scenarios' using markers such as 'IMPORTANT: This is a real scenario' and 'You must choose and act'. While these patterns are similar to those used in jailbreaking, here they are specifically employed as a methodology for testing AI compliance under simulated constraints like time and sunk cost.

writing-skills