The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to bypass security limits of external services (WeChat Official Accounts) by manipulating browser parameters like User-Agent and disable-blink-features.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection attacks. It retrieves content from arbitrary user-provided URLs through the web-article-extractor tool and processes the raw text for analysis.
Ingestion points: External URLs processed by the web-article-extractor tool (SKILL.md).
Boundary markers: None identified; extracted content is processed as a unified "super sample" string.
Capability inventory: Executes a bundled Python script (normalize_style_frontmatter.py) and performs file system writes to the .claude/styles/ directory.
Sanitization: None; the skill uses the raw extracted text directly for its 15-dimension analysis.
[COMMAND_EXECUTION]: The skill invokes a local Python script normalize_style_frontmatter.py to maintain and normalize YAML frontmatter in its style profile database. While the script itself performs safe regex-based text processing, the pattern of executing scripts on local files is noted.

style-modeler