The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to scrape and process content from external websites, which represents an indirect prompt injection surface. An attacker-controlled website could include malicious instructions intended to influence the agent's behavior. This risk is inherent to the skill's primary purpose of web scraping.
Ingestion points: Browser automation logic in SKILL.md and templates (basic-scrape.ts, stealth-mode.ts) ingests untrusted HTML and text data via page.goto and page.evaluate.
Boundary markers: The skill does not explicitly define boundary markers for untrusted data, but it focuses on structured extraction (JSON).
Capability inventory: The skill utilizes file writing (fs.writeFile in authenticated-session.ts), network operations (full browser access), and command execution capabilities (running automation scripts).
Sanitization: Standard Playwright extraction methods (textContent) are used; no additional sanitization is implemented within the provided templates.
[COMMAND_EXECUTION]: The skill provides templates and a shell script (install-browsers.sh) for environment setup and execution of automation tasks. These are standard operations for the tool's intended use in a developer environment.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of browser binaries and dependencies from official and well-known registries (NPM and PyPI). These include the core Playwright library and established community plugins for stealth and anti-detection.

playwright-local