The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a mandatory sanitization step for the <slug> identifier using a strict regular expression (^[a-z0-9][a-z0-9-]{0,39}$). This prevents path traversal and injection attacks during file operations.
[PROMPT_INJECTION]: Instructions emphasize developer-led decision-making and manual approval (HARD-GATE), avoiding any attempts to bypass or ignore agent safety guidelines.
[SAFE]: Tool usage is limited to local project exploration tasks, and the skill does not exhibit any network activity, obfuscation, or remote dependency risks.

brainstorming