The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill implements and provides extensive documentation for using the !command syntax to perform runtime environment probes. These probes (e.g., checking tool versions with command -v or verifying the presence of environment variables) are benign and serve the legitimate purpose of allowing the agent to adapt to the user's local setup. The skill acts as a developer guide for this platform feature.
[INDIRECT_PROMPT_INJECTION]: Because the skill generates new skill instructions based on user-provided requirements, it inherently possesses an ingestion surface for untrusted data. However, the skill employs structured templates, explicit defaults, and quality rubrics which help ensure that the generated output remains within safe operational boundaries.

skill-creator