The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/owl_parser.py and scripts/ontology_summarizer.py scripts allow fetching ontology data from remote URLs using urllib.request.urlopen. This is a standard feature for retrieving up-to-date scientific definitions from repositories like GitHub.
[PROMPT_INJECTION]: The skill processes descriptive metadata from external OWL/XML files, which creates a potential surface for indirect prompt injection.
Ingestion points: owl_parser.py fetches content from remote URLs or local file paths provided by the user.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potential instructions embedded within ontology labels or descriptions.
Capability inventory: The agent has access to Bash and Read tools, which could be targeted by successful injection.
Sanitization: The scripts parse the XML structure but do not sanitize the resulting text content for instructional patterns before presenting it to the agent.

ontology-explorer