The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is designed to ingest untrusted data from arbitrary URLs and provide it to the agent as markdown, creating a significant indirect injection surface.
Ingestion points: WebFetch, curl via bash, Playwright browser automation, and the Bright Data MCP (SKILL.md).
Boundary markers: No mention of boundary markers or instructions to the agent to ignore embedded commands within the fetched content.
Capability inventory: The skill possesses extensive capabilities including bash access, automated browser control, and network communication via multiple tiers (SKILL.md).
Sanitization: There is no evidence of content sanitization or escaping before the external data is interpolated into the agent's context.
Command Execution (HIGH): The skill explicitly uses the Bash Tool to execute curl with custom headers for 'Tier 2' scraping.
Evidence: "Tier 2: Customized Curl
Chrome-like browser headers to bypass basic bot detection" (SKILL.md).
Risk: If the URL or headers are constructed using unsanitized inputs from the user or previous scraping steps, it provides a direct path for command injection on the host system.
External Dependencies & Downloads (MEDIUM): The skill relies on external browser automation (Playwright) and a third-party scraping service (Bright Data).
Evidence: References to "Browser Automation
Full browser automation using Playwright" and "Bright Data MCP" (SKILL.md).
Risk: This increases the attack surface through external binary execution and data processing through non-vetted third-party endpoints.

brightdata