Use a test profile, synthetic data, and no production credentials. This guide is strategy and evaluation guidance, not managed implementation or support.
Start from the official documentation
Use the Hermes documentation and repository links as the source of truth. Check current platform support and installation instructions before running a command.
Install and run setup
Use the documented setup flow to choose a provider and confirm the agent starts. Keep credentials in the supported secret store; never commit them to a project.
Create a least-privilege profile
Use a separate profile for experiments. Start with the smallest toolset and no sensitive workspace so you can observe what the agent can read, write, execute, and send.
Test one read-only workflow
Run a research or documentation task first. Record tool calls, source quality, latency, and whether the output is reproducible.
Add capabilities one at a time
Enable terminal, browser, MCP, messaging, cron, or delegation only after the previous capability is understood. Review approvals and isolation settings before enabling actions.
Create a reset and exit plan
Document the profile location, provider, enabled tools, scheduled jobs, and how to revoke credentials and remove the profile.
Completion checklist
- Provider and model recorded
- Experimental profile isolated
- Tool permissions reviewed
- No secrets in repository
- Read-only task completed
- Reset path documented