The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content (web pages) which creates a potential surface for indirect prompt injection.\n
Ingestion points: The scripts/fetch_page.py script fetches and extracts text from arbitrary external URLs provided during runtime.\n
Boundary markers: The skill includes explicit instructions in SKILL.md and references/skill-contract.md to treat scraped material as evidence only and to ignore any embedded instructions found within that data.\n
Capability inventory: Capabilities include network reading via scripts/fetch_page.py and file writing via scripts/generate_markdown_report.py, scripts/generate_pdf_report.py, and scripts/schema_generator.py.\n
Sanitization: scripts/fetch_page.py performs basic HTML tag stripping and entity unescaping before the data is presented to the agent. scripts/schema_generator.py utilizes JSON-safe escaping when interpolating values into templates.\n- [EXTERNAL_DOWNLOADS]: The scripts/fetch_page.py script performs network requests to fetch web content.\n
Security Controls: The script implements protection against Server-Side Request Forgery (SSRF) by resolving hostnames and validating that the resulting IP address is global (public). It also rejects URLs containing embedded credentials (username/password).\n- [COMMAND_EXECUTION]: The skill includes several Python scripts designed to be executed by the agent to perform data fetching, SEO scoring, and report generation tasks. These scripts interact with the network and the local filesystem to produce artifacts.

seo-image-gen