The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted external content, creating a surface for indirect prompt injection.
Ingestion points: scripts/fetch_page.py is used to fetch and parse raw HTML content from public URLs.
Boundary markers: The SKILL.md and references/skill-contract.md files provide clear instructions to treat external evidence as untrusted input and explicitly warn against following embedded instructions.
Capability inventory: The skill has the ability to perform network requests, process data, and write reports to the local file system.
Sanitization: While the scripts extract specific metadata (title, H1, meta description) via regex, the agent is directed to process the broader page text, relying on instruction-based guardrails rather than strict programmatic sanitization.
[COMMAND_EXECUTION]: The skill executes several local Python scripts to perform its audit logic, which is the intended behavior of the skill.
The scripts include fetch_page.py for web requests, seo_health_score.py for weighted calculations, schema_generator.py for JSON-LD generation, and report renderers for Markdown and PDF formats.
scripts/fetch_page.py includes commendable security features such as hostname validation and IP address checks to prevent SSRF against private networks.

seo-audit