The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructions generate and execute short Python scripts via bash heredocs to extract site name, descriptions, and metrics from JSON and YAML files. These are used for internal data processing.
[COMMAND_EXECUTION]: The skill executes shell commands to manage its environment and assets, including npm install pptxgenjs, bk-capture-screenshot, and local utility scripts like bk-generate-book-metrics.
[EXTERNAL_DOWNLOADS]: In its press-release workflow, the skill uses curl to fetch the sitemap.xml from a user-provided URL. This is used as a fallback to verify project structure when the local repository is not available.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from local files such as about.md and mkdocs.yml to generate promotional text. This creates a standard surface for indirect prompt injection where untrusted content in those files could influence the agent's generated output.
[NO_CODE]: The standalone Python scripts provided (collect-site-metrics.py, crop-screenshot.py, validate-readme.py) are benign administrative and image processing tools that do not perform network operations or access sensitive system paths.

book-publisher