The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The owl_parser.py and ontology_summarizer.py scripts allow fetching ontology files from remote HTTPS sources. This is a core feature for accessing public materials science ontologies like CMSO and ASMO hosted on GitHub.
Evidence: scripts/owl_parser.py (line 214) uses urllib.request.urlopen with a 30-second timeout.
[COMMAND_EXECUTION]: The skill executes local Python scripts via Bash to process ontology data and return results to the agent. All subprocess calls use explicit argument lists.
Evidence: SKILL.md documentation provides explicit command-line examples for running class_browser.py, property_lookup.py, owl_parser.py, and ontology_summarizer.py.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of OWL/XML files, which could potentially contain malicious text designed to influence agent behavior.
Ingestion points: scripts/owl_parser.py reads data from a user-supplied file path or URL via the --source argument.
Boundary markers: The parser outputs structured JSON, providing clear separation between the data and the agent's instructions.
Capability inventory: The skill is limited to Bash and Read tools; scripts use Python's standard library only.
Sanitization: Input parameters like --class, --property, and --search are validated against a restricted safe-character regex (^[\w \-/()]{1,128}$) and length-capped at 128 characters.

ontology-explorer