orionbelt-ontology-builder

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md and references/upstream.md describe installing the orionbelt-ontology-builder package from a public registry using the uv tool. While the skill includes approval gates, this involves downloading and running code from an external source.
  • [COMMAND_EXECUTION]: The agent is instructed to use Bash to install and launch the workbench application (orionbelt-ontology-builder). The instructions explicitly require the agent to seek user approval before executing these commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and edit user-supplied ontology data (OWL/SKOS), which could potentially contain malicious instructions or malformed data designed to influence agent behavior.
  • Ingestion points: The agent uses the Read tool to process user-provided ontology files and local source code repositories.
  • Boundary markers: The instructions mandate previewing diffs, identifying target ontologies, and obtaining authorization before writing data or applying imports.
  • Capability inventory: The agent has access to shell execution via Bash and file system access through Read, Grep, and Glob tools.
  • Sanitization: The skill documentation suggests using review panels, change reports, and validation steps to manage the risks of importing external data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 09:02 AM