The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and fetches content from external, untrusted URLs (the target page and top search competitors) to perform its audit. This exposes the agent to indirect prompt injection, where an attacker could place malicious instructions on a web page to override the agent's analysis or behavior.
Ingestion points: SKILL.md (Step 1 and Step 3) specifies fetching and reading full rendered content from user-provided and search-discovered URLs.
Boundary markers: The instructions do not define clear delimiters or specific instructions to treat the fetched content strictly as data, increasing the risk of the agent following instructions found within the audited content.
Capability inventory: The skill utilizes URL fetching and Google search. It does not appear to use file-write or high-privilege shell capabilities.
Sanitization: There is no evidence of content sanitization or filtering to prevent the processing of malicious payloads embedded in HTML or markdown.
[DATA_EXFILTRATION]: The skill initiates network operations to arbitrary, non-whitelisted domains to fetch the target and competitor pages. While this is necessary for the skill's primary function (SEO auditing), these outbound requests to unknown servers constitute a minor data exposure risk in certain environments.

page-audit