update-nanoclaw
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches updates and script controllers from the vendor's official GitHub repository (nanocoai/nanoclaw) using git fetch and git archive.\n- [REMOTE_CODE_EXECUTION]: Extracts scripts from the fetched archive and executes them locally using tsx to manage the update, validation, and cutover logic.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests instructions from other installed SKILL.md files during the migration process, which could potentially contain malicious content.\n
- Ingestion points: Reads SKILL.md from referenced skills within the cut-over checkout (e.g., / references).\n
- Boundary markers: No explicit delimiters or instructions to ignore embedded prompts are mentioned for the ingested skill content.\n
- Capability inventory: The skill possesses significant capabilities including executing shell commands via tsx, managing system services (systemd, launchd), and modifying project files.\n
- Sanitization: The skill does not specify sanitization or validation of the instructions found within those external skill files.\n- [DATA_EXFILTRATION]: Accesses sensitive configuration files including .env, data/, and store/ to create snapshots before the update cutover. This is a local backup operation intended for safety and rollback functionality.
Audit Metadata