The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include a command for verification: node ${CLAUDE_SKILL_DIR}/../../scripts/psi-client.mjs --url <u>. This command incorporates a URL placeholder (<u>) that is intended to be replaced with a URL from the audited website. If the agent does not properly sanitize this input before execution, it could lead to command injection vulnerabilities.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks because it processes untrusted HTML and rendered DOM data from external websites.
Ingestion points: Untrusted data enters the agent's context through WebFetch, rendered_dom, and raw_html as described in SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to ignore or isolate potentially malicious commands embedded in the audited web content.
Capability inventory: The skill has access to the Bash tool (allowing shell command execution), WebFetch for network access, and file reading capabilities.
Sanitization: No sanitization, escaping, or validation of the external content is mentioned before it is analyzed or used in command parameters.
[EXTERNAL_DOWNLOADS]: The skill uses the WebFetch tool to download and analyze the content of arbitrary third-party websites. While this is the intended purpose of the SEO audit, it involves interacting with untrusted remote resources.

seo-core-web-vitals