The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves monitoring external web content, which acts as an ingestion point for untrusted data. If a monitored website is compromised or designed maliciously, it could serve prompts via HTML, console logs, or data streams that influence the agent's behavior during the verification process.
Ingestion points: External URL content (DOM elements, headers), browser console logs, and Server-Sent Event (SSE) streams as described in the 'What It Watches' section.
Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore embedded commands within the monitored data.
Capability inventory: The instructions imply the agent uses browser automation tools to inspect network responses, read console output, and extract page content.
Sanitization: There are no documented requirements for filtering or sanitizing the data retrieved from external URLs before the agent processes it.
[NO_CODE]: The provided skill file contains only Markdown documentation, configuration templates, and usage examples. No executable scripts, binaries, or automated installation routines are present in the skill directory, limiting the risk of direct malicious execution.

canary-watch