5dive-cli-extras
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONPERSISTENCEEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: Documents commands to grant agents unrestricted root access (
agent grant ... root) or modify system-wide sudoers configurations through managed templates. - [COMMAND_EXECUTION]: Provides an extensive interface for executing system commands, managing background units, and modifying host configurations (e.g.,
5dive host,5dive agent config). - [REMOTE_CODE_EXECUTION]: Facilitates the installation of external code via plugins from GitHub repositories and the import of agent packs that may contain arbitrary shell hooks.
- [DATA_EXFILTRATION]: Includes tools for managing sensitive credentials and describes features that transmit agent data to external AI providers (Grok/xAI) or vendor-hosted services (a2a.5dive.ai).
- [INDIRECT_PROMPT_INJECTION]: The skill manages the ingestion of untrusted external content. 1. Ingestion points: Agent packs (.tar.gz), company templates (team.yaml), and external event triggers (GitHub/webhooks). 2. Boundary markers: The CLI implements security tripwires for secrets and enforces deny-by-default for shell hooks unless explicitly allowed. 3. Capability inventory: The skill provides comprehensive system access including root privilege escalation, file writes, network operations, and plugin installation. 4. Sanitization: Includes zip-slip and link-escape guards for agent imports and secret-detection logic.
- [PERSISTENCE]: Enables the creation of recurring tasks and system-level timers/cron jobs for maintenance, status badges, and automated updates.
- [EXTERNAL_DOWNLOADS]: Manages downloads for system updates, plugins, and agent personas from external registries and repositories like GitHub.
Audit Metadata