The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze browser runtime state, including the DOM, console logs, and network payloads. This data originates from the website under test and represents an attack surface for indirect prompt injection if the target website contains instructions designed to influence the agent. The skill mitigates this by including explicit 'Do Not' and 'Stop If' instructions regarding sensitive data, authenticated storage, and unauthorized mutations.
[SAFE]: The skill provides defensive guidelines for developer workflows, such as warning against leaving development servers running and advising against the inspection of sensitive storage values unless specifically required by the task and allowed by privacy policy.

browser-runtime-testing