The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network operations to fetch data from external websites.
Evidence: The scripts/agentize_client.py script uses urllib.request.urlopen to connect to origins provided as arguments.
Context: The script includes safety checks such as validating that endpoints remain on the same origin and restricting response sizes to 8MB.
[PROMPT_INJECTION]: The skill processes untrusted data from external websites that could contain malicious instructions, creating an indirect injection surface.
Ingestion points: scripts/agentize_client.py retrieves external content in the command_search and command_get functions.
Boundary markers: SKILL.md does not instruct the agent to wrap the retrieved content in delimiters or use specific warnings to ignore embedded instructions.
Capability inventory: The agent executes a local Python script for network requests; no arbitrary command execution, file writing, or administrative privilege capabilities are observed.
Sanitization: The script performs structural validation of JSON and enforces origin matching, but does not sanitize the natural language content (Markdown) before providing it to the agent.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python helper script to interact with the protocol.
Evidence: SKILL.md provides multiple usage examples of python3 scripts/agentize_client.py for detection, search, and retrieval tasks.

search-agentized-websites