Hermes Command Cheat Sheet ฉบับอัปเดต — คำสั่งที่ควรรู้ไว้ใช้จริง

Hermes Command Cheat Sheet Update — Essential Commands to Know Now

This cheat sheet, prepared by Nokka (นก-กา), was written by AI (DeepSeek V4 Pro) through Hermes Agent and verified and edited by Nokka Hermes Agent. Recent command releases have left many users struggling to keep up, particularly regarding repetitive tasks and managing currently running tasks.

The cheat sheet compiles the most frequently used essential commands, organized into categories with instructions on when to use each one. It is based on @HermesWatcher's cheat sheet and cross-checked with Hermes source code.

1. Sessions & Context

- /new [name] — Start a new session (clears context)

- /resume [name] — Return to previously recorded session

- /branch [name] — Create a new branch path (experiment without destroying main path)

- /compress — Compress context to increase free space

- /context [all] — Check current context window usage

- /status — Overview: session, model, token, context

2. Control Active Work

This section is crucial for those using Hermes, as it details how to control agent workflows while they are active:

- /queue prompt — Add the next job to the queue without interrupting current work

- /steer prompt — Insert instructions after a tool call (adjusting midpoint direction)

- /bg prompt — Run a task in the background, separated from the main task

- /btw question — Pose a side question without hindering main work

- /agents — View active agents and tasks

- /stop — Stop all running tasks and their background jobs

3. Goals, Loops & Plans

Managing repetitive tasks and goals:

- /goal text — Set a long-term goal for Hermes to continue working towards

- /subgoal text — Add sub-goals to the active goal

- /heartbeat every interval prompt — Re-run the prompt when the session is idle

- /loop [interval] prompt — Repeat the task according to the interval, with stop conditions

- /plan [task] — Write a plan without executing it

- /review [instructions] — Create an independent agent to review recently completed tasks

- /refine [focus] — Revisit the conversation, noting lessons learned in memory/skills

4. Models & Behavior

Switching models and adjusting behavior:

- /model [model] — Switch to a different model

- /moa prompt — Run a prompt through the Mixture of Agents preset and return to the original model

- /personality [name] — Set or clear personality overlay

- /reasoning ... — Set the reasoning level for generating evidence

- /fast ... — Control the speed mode

- /priority ... — Set the priority level of risky commands

- /yolo — Switch to bypass risky command approvals

- /busy [queue|steer|interrupt|status] — Adjust behavior when new messages arrive while busy

- /voice ... — Control voice settings

5. Skills, Memory & Tools

Managing skills, memory, and tools:

- /skills ... — Search, install, or manage skills

- /learn source — Change the workflow or source to a reusable skill

- /memory ... — Check or approve memory writing

- /init [notes] — Create or update AGENTS.md from the repository

- /tools ... — List, open, or close tools

- /browser ... — Connect or manage browser tools

6. Automation & Coordination

Automation and scheduling:

- /cron ... — Schedule durable off-chat tasks

- /suggestions ... — Review system-proposed automation

- /blueprint ... — Start automation from blueprint templates

- /kanban ... — Manage collaborative work boards across profiles

7. Inspect, Recover & Fix

Checking, recovering, and fixing issues:

- /retry — Retry the last message

- /undo [N] — Reverse the last N changes

- /save — Save or export the conversation

- /diff ... — View Git changes in the working directory

- /rollback [N] — Show or restore a filesystem checkpoint

- /usage — View token usage and limitations

- /debug [nous|local] — Generate a support report

- /help [filter] — Display or edit commands, helpful when users are confused

Key points to remember:

The most frequently used commands fall into two categories:

- Control Active Work (/queue, /steer, /bg, /btw)

- Goals & Loops (/goal, /loop, /heartbeat)

Focus on these four commands first:

- /steer (adjusting agent direction),

- /bg (run tasks in the background),

- /loop (repeat within a session),

- /cron (repeat regardless of chat closure)

To summarize: if you want a task to persist even when the chat is closed, use /cron. If you want the action to run within the current session, use /loop. If you want the work to continue until the goal is achieved, use /goal. If you want Hermes to check the progress periodically while idle, use /heartbeat.