The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The preflight.py script includes instructions for installing the Google Cloud SDK using a remote script execution pattern (curl https://sdk.cloud.google.com | bash). This download targets an official Google domain, which is a recognized trusted source, and is intended for environmental setup.
[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with the gcloud CLI for authentication and API management. This functionality is restricted to specific gcloud subcommands and is a core requirement for fetching search performance data.
[PROMPT_INJECTION]: The skill processes untrusted data from external website crawls and CMS APIs (including WordPress, Strapi, Ghost, and Contentful) to generate reports and content recommendations, creating a surface for indirect prompt injection.
Ingestion points: Website metadata and body content (Phase 5 technical crawl) and CMS article data (Phase 3.6).
Boundary markers: The skill uses structured JSON parsing to separate data fields, though it does not implement explicit textual delimiters when passing external content to secondary agents.
Capability inventory: The skill can execute shell commands (limited to gcloud), perform network requests to trusted APIs, and write audit logs to the local filesystem.
Sanitization: Data is parsed via standard libraries, but the raw text is passed to the LLM for analysis without specific injection-focused filtering.

seo-analysis