The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/citability_scorer.py utilizes __import__('sys') to dynamically load the system module for reading input from stdin. This is a common pattern in standalone scripts to avoid global imports when module functionality is only required in a specific conditional branch or one-liner.- [INDIRECT_PROMPT_INJECTION]: The skill processes external data, creating an attack surface. 1. Ingestion points: scripts/fetch_page.py (URL fetching) and scripts/brand_scanner.py (local file scanning). 2. Boundary markers: Explicit warnings in SKILL.md and references/skill-contract.md instructing the agent to treat evidence as untrusted input. 3. Capability inventory: Network requests in scripts/fetch_page.py, file writes in scripts/generate_markdown_report.py and scripts/generate_pdf_report.py. 4. Sanitization: Instruction-based boundaries directing the agent to ignore instructions embedded in scraped data.- [EXTERNAL_DOWNLOADS]: The scripts/fetch_page.py script executes network requests to fetch web pages. The script implements security controls, including hostname validation and SSRF protection (blocking private, loopback, and reserved addresses) to prevent unauthorized access to internal network resources.

geo-prospect