Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/liustack/summono/agents-mdgit clone --depth 1 https://github.com/liustack/summonoWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00883 | $0.00883 |
| Opus 5 | $0.00441 | $0.00441 |
| Sonnet 5 | $0.00177 | $0.00177 |
| Haiku 4.5 | $0.00088 | $0.00088 |
Grade A, and why
summono AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Project Overview (for AI Agent)
Goal
Summono is the installer and launcher for AI harnesses — a desktop app that installs and starts an AI agent in one click, for people who never open a terminal. Free and open source forever (MIT).
Two shapes of harness, two treatments:
- Harnesses with their own desktop GUI (Codex app, Claude Desktop): install only, never wrap — download and install the official app plus whatever it needs, then launch it from the panel.
- Harnesses with no desktop form (DeepSeek Harness): Summono supplies an invisible runtime, hosts the process on a dedicated port, and embeds the web UI inside its own window.
Technical Approach
- Three processes, one hard boundary. All real work lives in a resident engine process (
utilityProcess); the Electron main process only manages windows, supervises the engine, and relays IPC. Engine ↔ GUI is pure JSON-RPC ({id,method,params}/{id,result|error}, events as{event,payload}), so the engine can be swapped out wholesale without touching the app shell. - Vendor passthrough, never replacement. Whatever a vendor already ships — official installers, desktop apps, state-directory conventions (dsh's
~/.dsh) — is used as-is. Summono only fills the gaps vendors don't cover. - Hosted instances are process-isolated but state-shared. The hosted dsh runs on its own port (34517) but reuses the vendor's default state directory, so hand-installed copies coexist and every online tutorial still applies.
- Private runtime supply. Node.js is downloaded into Summono's own directory and never touches the user's system; all platform differences converge in
runtime.ts. - Organized by business domain, not by type. A new capability starts with "which domain does this belong to?" — never a
handlers/orutils/directory.
Code Organization
src/
├── main/ # Electron main process: window management, engine supervision, IPC relay — no business logic
├── engine/ # resident engine process, one file per business domain
│ ├── runtime.ts # private Node.js supply, all platform differences
│ ├── sources.ts # artifact fetch: official + mirror fallback, stall, sha256
│ ├── dsh.ts # dsh: install, launch, probe, supervise
│ └── apps.ts # desktop apps: detect and launch
├── preload/ # contextBridge — wiring only
└── renderer/ # the React launchpad
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 57 lines · 883 tokens per session scan A f6c622a9dbf0
summono AGENTS.md is an instructions file published in the GitHub repository liustack/summono (2 stars, last pushed 9d ago), licensed MIT. It adds 883 tokens to every session, about $0.0044 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
deepseek-harness-desktop AGENTS.md
Instructions for baihejiangnan/deepseek-harness-desktop, covering dsh launcher 开发规范, 首要原则, 进程与窗口模型, 实例与数据规则 and dsh 运行时与更新边界.
desktop-cc-gui AGENTS.md
AGENTS.md instructions for zhukunpenglinyutong/desktop-cc-gui, covering 项目规则入口（mossx）, 规则优先级, 文档分层, 最小读取路径 and openspec 交付.
DSH-better-sidebar AGENTS.md
Instructions for omdsh-dev/DSH-better-sidebar, covering dsh-better-sidebar 仓库规则（agents）, 1. 仓库硬约束（必须遵守）, 2. ci 挂载冒烟（plugin-mount job / pnpm test:mount）, 3. dsh 0.1.2-alpha 适配要点（alpha 通道） and 4. npm 发版（github release → npm publish）.
dsh-worktable AGENTS.md
Instructions for Aisland-SJL/dsh-worktable, covering dsh-worktable 项目规则, 协作方式（用户定案，最高优先级）, 边界, 构建与验证 and 领域约定（会话中必须遵守）.
DeepSeek-Harness-Desktop AGENTS.md
Instructions for web-casa/DeepSeek-Harness-Desktop, covering agents.md — dsh desktop, 项目本质, 不可违背的安全不变量, 版本单一事实源（改动版本时必须同时改全部） and harness 升级启动契约（bump pin 前必读）.
awesome-deepseek-harness-plugins AGENTS.md
Instructions for imsai-sh/awesome-deepseek-harness-plugins, covering repository instructions, the submission gate is the product, generated files, cross-repo contracts (no ci spans both repositories) and permanent urls.