worker-management
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes functionality to import and install custom Worker Skills from ZIP archives through
scripts/install-worker-skill.sh. This processes external data that could contain malicious instructions for the worker agent. - Ingestion points: ZIP archives processed via the
--archiveflag inscripts/install-worker-skill.sh. - Boundary markers: The
SOUL.mdtemplate inreferences/create-worker.mdincludes explicit 'Security Rules' intended to guide the agent's behavior. - Capability inventory: The skill utilizes the
agtCLI for worker orchestration,mc(MinIO) for storage operations, andcurlfor Matrix communication. - Sanitization: A dedicated validator (
scripts/safe-extract-worker-skill.py) is used to prevent Zip Slip (path traversal), symlinks, and archive bomb attacks. - [COMMAND_EXECUTION]: The skill makes extensive use of the
agtCLI tool for managing workers and teams. It also executesmcfor mirroring configurations to object storage andcurlfor Matrix API interactions. These commands are integral to its purpose as a management tool. - [EXTERNAL_DOWNLOADS]: The skill is designed to handle local ZIP attachments provided by an administrator to install or update worker capabilities. The installation process includes verification of the archive's structure and metadata.
Audit Metadata