The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents browser automation which introduces a surface for indirect prompt injection. A malicious website could attempt to influence the agent's behavior through the content displayed in the browser during an automated session. \n
Ingestion points: page.goto(url) in SKILL.md and reference.md. \n
Boundary markers: None present in the provided templates to isolate browser content from instructions. \n
Capability inventory: Use of page.evaluate(), fs.renameSync(), and execSync() for video processing. \n
Sanitization: The templates do not show sanitization of external content before processing. \n- [COMMAND_EXECUTION]: The documentation suggests the use of shell commands for installing dependencies, running tests, and processing video files. \n
Evidence: execSync used with ffprobe in reference.md; suggested ffmpeg and npx commands in SKILL.md. \n- [EXTERNAL_DOWNLOADS]: The skill instructs the user to download and install external dependencies and browser binaries. \n
Evidence: npm install -D playwright @playwright/test and npx playwright install chromium.

playwright-recording