The Agent Skills Directory

[SAFE]: The skill provides utility scripts (init_skill.py, validate_skill.py, package_skill.py) that perform standard file system operations for skill management. These tools are used for initializing directory structures, validating YAML frontmatter, and packaging skills without any detected malicious behavior.
[DYNAMIC_CONTEXT_INJECTION]: The skill documents the use of dynamic context injection via the !command syntax to fetch live environment data. The provided examples, such as git branch and git diff, are benign development tasks and represent legitimate usage of the platform's capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a generator for new instructions based on user input, which presents an attack surface for indirect prompt injection. Ingestion points: Untrusted data is ingested from $ARGUMENTS and responses to the AskUserQuestion tool in SKILL.md. Boundary markers: The skill does not explicitly apply boundary markers to the generated output, although it provides strict standards for description density and instruction structure. Capability inventory: The skill performs file and directory creation, structural validation, and ZIP-based packaging through its included Python scripts. Sanitization: The init_skill.py script normalizes skill names to hyphen-case using regular expressions, though the semantic content of the generated instructions is derived from user requirements without further sanitization.

create-skill