The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses Playwright to automate browser sessions, which involves visiting and interacting with arbitrary external websites. This creates a surface for indirect prompt injection where malicious instructions embedded in a website's content could influence the agent's actions.
Ingestion points: page.goto(), page.click(), and page.fill() operations across SKILL.md and reference.md are used to process content from external URLs.
Boundary markers: No specific instructions are provided to the agent to ignore or delimit instructions found within the processed web content.
Capability inventory: The skill possesses significant capabilities, including file system access (fs.renameSync, fs.mkdirSync), network operations (browser navigation), and shell command execution (execSync).
Sanitization: The skill does not explicitly describe sanitization or filtering of text extracted from web pages before it might be processed by an LLM.
[COMMAND_EXECUTION]: The skill documentation and reference material include instructions for executing shell commands, specifically ffmpeg for video conversion and ffprobe (via child_process.execSync) for extracting video metadata.
[EXTERNAL_DOWNLOADS]: The skill guides the user to install the Playwright library and its browser binaries (npx playwright install chromium), which involves downloading executable content from the Playwright project's servers.

playwright-recording