The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses sudo to execute the 5dive CLI for system-level management tasks, including creating Linux users and systemd units.\n
Evidence: sudo 5dive agent create, sudo 5dive agent rm in SKILL.md.\n- [CREDENTIALS_UNSAFE]: The skill manages environment files and API keys in privileged directory locations.\n
Evidence: /etc/5dive/connectors/ and /var/lib/5dive/agents.d/ in references/paths.md.\n
Evidence: Use of auth set --api-key=- to provision credentials.\n- [PROMPT_INJECTION]: The skill interpolates user-supplied text directly into shell commands for inter-agent communication, creating a surface for command injection or indirect instruction injection.\n
Ingestion points: User prompts and side-task descriptions in SKILL.md.\n
Boundary markers: None present in the provided shell recipes.\n
Capability inventory: High-privilege shell access via sudo across all scripts.\n
Sanitization: No explicit escaping or validation of the prompt variable before interpolation into the 5dive agent send command.\n- [EXTERNAL_DOWNLOADS]: The skill supports fetching additional tools or instructions from external repositories during agent provisioning.\n
Evidence: The --with-skills flag in references/commands.md facilitates remote resource fetching.

5dive-cli