The Agent Skills Directory

[PROMPT_INJECTION]: The skill explicitly instructs on the use of 'persuasion principles' (Authority, Scarcity, Commitment, Social Proof, and Unity) to maximize agent compliance and minimize rationalization. It utilizes strong language markers like 'THE IRON LAW', 'HARD-GATE', and 'YOU MUST' to enforce adherence to its internal methodology.- [COMMAND_EXECUTION]: Phase 4 of the main skill body provides instructions for executing local scripts (quick_validate.py, check-markdown-links.sh, and sync-skills.sh) via python3 and bash. These scripts are located in local system paths (e.g., $HOME/.codex) and are used for skill verification.- [PROMPT_INJECTION]: The skill features a 'Pressure Test' methodology designed to stress-test an agent's willingness to follow rules by simulating high-stakes environments, such as time constraints, authority overrides, and economic pressure. This is a form of adversarial prompt engineering designed to strengthen rule adherence.- [PROMPT_INJECTION]: A potential indirect injection surface is present in the shell command templates in Phase 4 of SKILL.md.
Ingestion points: The <skill-name> placeholder in the validation commands in SKILL.md.
Boundary markers: No delimiters or safety warnings are provided for the placeholders within the shell commands.
Capability inventory: The skill utilizes python3 and bash for local script execution.
Sanitization: There is no explicit instruction or mechanism for sanitizing or escaping the <skill-name> input before it is interpolated into the shell environment.

writing-khuym-skills