The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script scripts/docs-seo-crawl.py to perform documentation site audits. This involves running shell commands to fetch and analyze URLs and sitemaps.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it fetches and processes content from external, potentially untrusted documentation websites, which is then analyzed by the agent.
Ingestion points: The scripts/docs-seo-crawl.py script fetches HTML and XML content from external URLs and sitemaps provided by the user or discovered during the audit (in SKILL.md Step 1).
Boundary markers: Audit results are stored in structured csv and json formats (e.g., pages.csv). However, the agent is instructed to read these results and generate a "verdict paragraph" and a "ranked fix list," which involves processing the ingested external text.
Capability inventory: The skill executes local Python scripts, performs network GET requests to user-provided URLs, and writes results to local project files like pages.csv, pages.json, and devrel-context.md (in SKILL.md Step 1 and Step 7).
Sanitization: The Python script performs basic cleaning by removing specific HTML tags (script, style, nav, etc.) and stripping other tags using regex, but it explicitly extracts and preserves text from titles, headings, and meta descriptions, which are then presented to the AI agent without further sanitization.

docs-seo