The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection vulnerability surface detected. Untrusted web content is ingested into the agent context, creating a potential vector for malicious instructions embedded in scraped pages to affect the agent's logic.
Ingestion points: Web content is read via document.body.innerText in both scripts/playwright-simple.js (line 42) and scripts/playwright-stealth.js (line 117).
Boundary markers: There are no explicit delimiters (like XML tags or markdown blocks) or instructions to the agent to ignore commands within the scraped content.
Capability inventory: The skill environment allows for file creation (fs.writeFileSync, page.screenshot), shell command execution (node, bash), and network access via Playwright.
Sanitization: The scraped content is processed as raw text and returned as a JSON object to the agent without any filtering, escaping, or validation.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright npm package and associated Chromium browser binaries as documented in INSTALL.md and SKILL.md. These downloads originate from official and well-known registries necessary for the skill's core functionality.
[COMMAND_EXECUTION]: The skill relies on executing Node.js scripts (scripts/playwright-simple.js, scripts/playwright-stealth.js) and shell scripts (examples/discuss-hk.sh) to perform its tasks. While these are legitimate for a scraper, they represent an execution surface that depends on the safety of the URLs provided as arguments.

playwright-scraper-skill