The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external websites and project files. It includes specific defensive instructions to mitigate this risk.
Ingestion points: Web page content fetched via scripts/fetch_page.py, project files, and third-party exports.
Boundary markers: SKILL.md and references/skill-contract.md explicitly instruct the agent to treat scraped material as untrusted input and ignore embedded instructions.
Capability inventory: The skill can perform network requests (fetch_page.py), write files (generate_markdown_report.py, generate_pdf_report.py, schema_generator.py), and execute local scripts.
Sanitization: Relies on instructional boundaries and explicit agent warnings rather than technical filtering.
[EXTERNAL_DOWNLOADS]: The script scripts/fetch_page.py allows the agent to download content from arbitrary URLs.
Evidence: Uses urllib.request to fetch content. It includes an SSRF protection mechanism that checks if IP addresses are global, although this check can be bypassed using the --allow-private flag.
[COMMAND_EXECUTION]: The skill provides a suite of Python scripts (fetch_page.py, generate_markdown_report.py, generate_pdf_report.py, schema_generator.py, seo_health_score.py) intended to be executed by the agent to perform its tasks.
[DYNAMIC_EXECUTION]: The script scripts/generate_pdf_report.py uses dynamic imports for the reportlab library and suggests external installation if missing. The scripts/schema_generator.py script performs string replacement on templates before parsing them with json.loads, which is a form of dynamic data assembly.

seo-geo