Skip to main content

code-mate-hermes

Run Hermes Agent non-interactively for local coding tasks. Use when the user asks to delegate work to Hermes.

Source facts

Repository
CherryHQ/cherry-studio
Last source activity
October 20, 2026 at 02:55
Detected SKILL.md language
English
Stars
53,185
Forks
5,159

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
code-mate-hermes
description
Run Hermes Agent non-interactively for local coding tasks. Use when the user asks to delegate work to Hermes.
# Hermes Agent ## Run 1. Set the Bash working directory to the smallest directory the user authorized and set a finite timeout, normally 10 minutes. 2. Check availability with `command -v hermes`. If it is missing, stop and ask the user to install Hermes Agent in Code Mate. 3. Run one task and exit: ```bash hermes -z "<prompt>" ``` Pass the prompt as one quoted argument. Treat stdout as final text. Exit zero means success, exit one means failure or no response, and exit two means invalid arguments or an empty failed/partial result. Never start the interactive UI or a login flow. ## Authentication And Permissions If Hermes reports missing provider, model, or API configuration, stop and ask the user to configure Hermes Agent in Code Mate. Never request, read, print, or copy credentials. One-shot mode automatically enables YOLO and hook acceptance and can load tools, memory, rules, and `AGENTS.md`. Do not run it in a writable project for read-only work. Use a disposable or read-only copy instead. Run it against the real project only when the user explicitly requests workspace changes, constrain the directory and prompt, and inspect the diff afterward. Example: for an approved edit, ask Hermes to touch only named files, run it in that project, require exit zero, and inspect the resulting diff before reporting success.
View on GitHub