The Agent Skills Directory

[SAFE]: The skill is designed to enforce coding style guidelines regarding documentation. It utilizes a limited set of non-destructive tools (Read, Glob, Grep) and does not request network access or write permissions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external content (source code comments) which could theoretically contain malicious instructions. However, the risk is addressed by the restricted tool capabilities and the specific instructional scope of the skill.
Ingestion points: The skill reads source code via the Read tool to audit comments.
Boundary markers: No explicit delimiters or boundary warnings are provided for the agent when reading files.
Capability inventory: The skill is restricted to Read, Glob, and Grep tools; it lacks the ability to execute commands, write files, or perform network operations.
Sanitization: No sanitization logic for ingested code is present.
[METADATA_POISONING]: The metadata field includes fictional model versions (e.g., gpt-5.6, gemini-2.5). While these are technically deceptive as they do not exist, they do not facilitate security bypasses or hide executable code and are assessed as benign stylistic choices.

wk-workstyle-docs