The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions require the agent to verify empirical facts by executing code. Specifically, it tells the agent to "Run the snippet, drive the browser, print the event order, check the actual output" in the SKILL.md file.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from both external web sources and the local codebase, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The agent uses WebSearch and WebFetch for external research and grep for local codebase analysis (SKILL.md).
Boundary markers: There are no instructions provided to treat the fetched data as untrusted or to use delimiters to prevent instruction leakage.
Capability inventory: The agent has the ability to write files to the local filesystem (e.g., /tmp/), perform network searches, and execute code snippets for verification.
Sanitization: No sanitization or validation steps are defined for the data retrieved from external or local sources.
[EXTERNAL_DOWNLOADS]: The skill uses WebSearch and WebFetch tools to retrieve technical specifications and documentation from external sites such as MDN, WHATWG, and RFC repositories during the mandatory research phase.

learn-html