Hermes Talaria is a command-line setup tool that creates a lean, profile-local Hermes Agent configuration with fewer active skills and lower token use. It copies your global skills into a new profile, keeps a minimal core set active, moves the rest into profile-local skills-heavy or skills-disabled folders, and patches the profile config into a light mode. Global ~/.hermes/skills is never modified.
The active skill set is limited to autonomous-ai-agents/hermes-agent, github/, mcp/native-mcp, and software-development/. The generated config sets toolsets: [], agent.reasoning_effort: low, disables the hermes-cli toolset, checkpoints, memory, user profiles, delegation, the curator, speech-to-text, the model catalog, and Honcho, and sets a concise display personality with a 60-minute prompt cache TTL. The run finishes by launching model to trigger Hermes model setup for the new profile.
Features
- Profile-local skills: global skills are copied into ~/.hermes/profiles/ /skills, leaving the global directory untouched
- Minimal core set: only the Hermes agent, GitHub, native MCP, and software-development skills stay active
- Heavy and disabled tiers: all other skills move into profile-local skills-heavy or skills-disabled
- Light-mode config: a patched config.yaml with low reasoning effort, no memory, no delegation, no checkpoints, and a concise personality
- Custom profile name: -n MyProfile targets a profile of your choosing
- Dry run: --dry-run previews the changes without writing them
- Model setup control: --no-model-setup skips the final model setup launch, which you can run later with model
