The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to navigate to external websites using the Playwright MCP server and extract data patterns to generate scraping scripts. This creates a surface where malicious content on a target website could influence the code generated by the agent.
Ingestion points: The skill uses browser_navigate and browser_snapshot (Step 4 in SKILL.md) to explore untrusted external web content.
Capability inventory: The skill has access to Bash, Write, and Edit tools, allowing it to create and execute code.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat scraped content as data rather than instructions during the script generation phase.
Sanitization: There is no evidence of sanitization or filtering of external content before it is interpolated into the script generation prompts.
[DYNAMIC_EXECUTION]: The skill generates and saves new Python or Node.js scripts (e.g., scrapers/scraper_name.py) based on templates in the references/ directory and then executes them using the Bash tool to verify functionality. This script generation and execution workflow happens at runtime based on the target website's structure.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to install patchright (via pip, npm, or npx), which is an unofficial community-maintained fork of Playwright. It also recommends using the webshare CLI tool for proxy management.
[COMMAND_EXECUTION]: The skill uses the Bash tool to install dependencies, run the generated scraper scripts, and interact with the webshare CLI. It also executes the scripts/init_project.py script to scaffold new projects.

scraper