nvidia-ontology-setup
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for executing system commands including
helm,docker compose,kubectl,uv, andpnpm. These are used to deploy, configure, and troubleshoot the Auto Ontology services as intended by the skill's purpose. - [EXTERNAL_DOWNLOADS]: Fetches deployment assets and software packages from trusted vendor sources, specifically the NVIDIA Helm registry (
helm.ngc.nvidia.com) and the official NVIDIA GitHub organization. - [INDIRECT_PROMPT_INJECTION]: The skill manages the connection and ingestion of external database content. While this creates a pathway for untrusted data to enter the agent's context, the skill provides documentation for scope-based filtering (allow/deny lists) to mitigate unauthorized data access.
- Ingestion points: Database connection registration via
POST /api/connections(SKILL.md). - Boundary markers: Use of
table_allow_regexandtable_deny_regexto define data scope (SKILL.md). - Capability inventory: Shell execution (
helm,docker,uv,pnpm), file system writes for configuration, and network requests to verified deployment endpoints. - Sanitization: Catalog extraction filters and status verification checks (
GET /api/semantic-compilation/status). - [SAFE]: The skill enforces a "no-secrets-in-chat" policy, instructing users to write sensitive credentials like API keys and database passwords directly to local configuration files (
.env,values.yaml) to prevent exposure in logs or shell history.
Audit Metadata