The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (detect_projects.py, feedback_state.py) using uv run to manage project state and feedback loops. It also instructs the agent to run standard development tools like scrapy and pytest within the project environment.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data by processing arbitrary target URLs and scraping their content to generate code.
Ingestion points: Target URL from $ARGUMENTS and HTML content processed via the /scrape-plan sub-skill.
Boundary markers: Relies on the planning stage and LLM-based validation before code generation.
Capability inventory: Writing project files (spiders, items, page objects) and executing them via uv run.
Sanitization: Relies on the orchestrated workflow to validate extraction logic before implementation.
[EXTERNAL_DOWNLOADS]: The skill fetches markdown documentation from well-known and trusted services, including docs.zyte.com, docs.scrapy.org, and readthedocs.io. These sources are used to provide the agent with library-specific context and best practices.

scrape