ontology-builder-assistant
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and normalize untrusted data from various formats (CSV, JSON, XML, YAML, and natural language documents) to derive ontology elements.\n
- Ingestion points: Workflow Step 1 in SKILL.md specifies the collection of "representative data" and "supporting semantic evidence" from external sources.\n
- Boundary markers: No specific delimiters or "ignore embedded instructions" directives are provided to protect the agent from instructions hidden within the input data.\n
- Capability inventory: The skill possesses the capability to write files (.ttl, .json) and execute a Python script (scripts/owl_to_graphrag_schema.py).\n
- Sanitization: No sanitization or validation logic is defined for the input data prior to processing.\n- [DYNAMIC_EXECUTION]: The skill involves the execution of a bundled Python script on generated data derived from user inputs.\n
- Evidence: Section 5 of the 'Output contract' in SKILL.md explicitly instructs the agent to "run the bundled conversion script scripts/owl_to_graphrag_schema.py" after generating the ontology.\n- [COMMAND_EXECUTION]: The skill directs the agent to invoke shell commands to run the conversion script, targeting files created during the session.
Audit Metadata