The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external websites (HTML source, robots.txt, sitemaps) using scripts like scripts/fetch_page.py and scripts/parse_html.py. This data is then analyzed by the AI agent to produce SEO reports. This creates a surface for indirect prompt injection if the fetched content contains malicious instructions designed to influence the agent's behavior during analysis.
[COMMAND_EXECUTION]: Several scripts, including scripts/generate_report.py, scripts/github_api.py, and scripts/lighthouse_runner.py, use the subprocess module to execute internal Python scripts or standard command-line tools like git, gh (GitHub CLI), and lighthouse. These commands are constructed using hardcoded paths and sanitized parameters (e.g., normalized repository slugs), representing controlled functionality essential to the skill's SEO auditing purpose.
[EXTERNAL_DOWNLOADS]: The skill performs legitimate network operations to fetch SEO data from well-known and trusted services, including Google (PageSpeed Insights, Autocomplete, Knowledge Graph), Wikipedia, Wikidata, and various search engine IndexNow endpoints. These requests are mediated by a secure HTTP wrapper (scripts/lib/safe_http.py) that implements SSRF guards by blocking private/internal IP addresses and enforcing response size limits.
[CREDENTIALS_UNSAFE]: The skill includes a dedicated environment loader (scripts/env_loader.py) and specific logic in scripts like scripts/pagespeed.py to handle API keys and tokens. These are loaded from system environment variables or local .env files, which is a standard and recommended practice for avoiding hardcoded secrets while allowing the use of authenticated APIs.

seo