nabu
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill performs a version check by fetching a remote configuration file from the author's repository at
https://raw.githubusercontent.com/Timmyy3000/skills/main/skills/nabu/SKILL.md. This is documented as a safety preflight to ensure the skill remains current. - [COMMAND_EXECUTION]: The skill includes instructions to perform updates using the
npx skills update nabucommand. This execution is gated by explicit user approval and is limited to updating the skill package itself. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Nabu deployments, including note contents and vault metadata.
- Ingestion points: Fetches version metadata from GitHub, deployment contracts from
${NABU_URL}/agents.md, and knowledge content via note search and read tools. - Boundary markers: The skill contains specific instructions to treat remote responses as untrusted data and explicitly warns against following instructions contained within those responses.
- Capability inventory: The skill can perform file system mutations (creating, updating, moving, and deleting notes) and make network requests via MCP or HTTP fallback API.
- Sanitization: Implements strict URL normalization, path validation (rejecting traversal, symlinks, and root paths), and mandatory human confirmation for shared space creation.
Audit Metadata