The Agent Skills Directory

[SAFE]: The skill's primary function is to deconstruct text styles across 15 dimensions (persona, logic, vocabulary, etc.) and save these as markdown files. All identified operations, such as saving files to docs/ or .claude/styles/, are consistent with its stated purpose of building a style library.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use a web-article-extractor tool when provided with URLs (specifically mentioning WeChat Official Accounts). This is a functional requirement for analyzing online content and does not involve downloading or executing untrusted code directly within the skill's logic.
[COMMAND_EXECUTION]: While the skill mentions configuring parameters like User-Agent and disable-blink-features to bypass anti-scraping measures on WeChat, these are instructions for the extraction tool's configuration rather than arbitrary shell command execution.
[DATA_EXFILTRATION]: No evidence of data exfiltration was found. The skill processes user-provided text or URLs and stores the analysis locally within the environment's project structure (docs/ and .claude/styles/).
[PROMPT_INJECTION]: The instructions do not contain patterns aimed at bypassing AI safety guidelines or overriding system prompts. It uses "IMPORTANT"-style markers for procedural accuracy (e.g., serial extraction vs. concurrent) which is benign instructional language.

style-modeler