The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses WebSearch and WebFetch to retrieve HTML content from arbitrary external URLs for competitor research and website auditing. This introduces untrusted data into the agent's processing environment.
[COMMAND_EXECUTION]: The skill uses the open command to launch the user's default browser to view the generated SEO reports on the local filesystem.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality.
Ingestion points: Retrieves content from external competitor websites, user-provided URLs, and local HTML directories.
Boundary markers: The instructions do not define clear delimiters or warnings for the agent to ignore instructions embedded within the fetched HTML content.
Capability inventory: The skill can perform web searches, fetch remote URLs, read and write local files, and execute the open shell command.
Sanitization: There are no explicit instructions to sanitize or filter potentially malicious elements (like <script> tags) from the retrieved content before it is interpolated into the HTML report templates.
[DYNAMIC_EXECUTION]: The skill generates local HTML files that include content derived from the analyzed websites. If the agent verbatim includes malicious code found on an analyzed site, it could lead to a localized Cross-Site Scripting (XSS) attack when the user views the report.

seo-strategy