The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from external URLs provided in arguments, which creates a surface for indirect prompt injection attacks where malicious instructions on a scanned page could influence the agent's behavior.
Ingestion points: Untrusted content is retrieved from external websites using the WebFetch tool based on the [url] argument.
Boundary markers: There are no specified delimiters or instructions (e.g., 'ignore embedded commands') provided to the agent to isolate the audited content from its system instructions.
Capability inventory: The skill utilizes Bash, WebFetch, Read, Grep, and Glob, providing a significant execution surface if an injection occurs.
Sanitization: The instructions lack any specific validation or sanitization steps for the data retrieved from the web before it is analyzed.

seo-technical