The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the geo command-line interface (CLI) via Bash to perform web audits and generate files. This involves executing commands that take external URLs as arguments and can write results to the local filesystem (e.g., geo llmstxt <url> --generate --out llms.txt).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites that may contain malicious instructions designed to subvert AI behavior.
Ingestion points: Data returned by the geo tool from crawled web pages, specifically within the findings[].excerpt and signals[].detail fields.
Boundary markers: The skill contains a 'Response contract' that explicitly instructs the agent to treat these excerpts as untrusted quotations and to ignore any embedded commands or instructions.
Capability inventory: Access to Bash for CLI execution and file writing, and Read for file access.
Sanitization: The skill includes a dedicated detection mechanism for content.instruction_like patterns and mandates reporting them as findings rather than following them.
[EXTERNAL_DOWNLOADS]: The instructions recommend the installation and upgrade of the seomator-geo-audit package via uv or pipx. This tool is a resource provided by the vendor to support the skill's functionality.

llmstxt