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/jacklandrin/onlyswitch/agents-mdgit clone --depth 1 https://github.com/jacklandrin/OnlySwitchWhat 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 | $0.01186 | $0.01186 |
| Opus 5 | $0.00593 | $0.00593 |
| Sonnet 5 | $0.00237 | $0.00237 |
| Haiku 4.5 | $0.00119 | $0.00119 |
Grade A, and why
OnlySwitch 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 2d ago.
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 — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.
OnlySwitch Agent Guidance
This file defines the default workflow for changes to the OnlySwitch project. It applies to the macOS app, widgets, remote app, Swift Package modules, tests, OpenClaw integration, and project documentation.
Workflow
For every task, follow these steps in order. Each phase is delegated to the named subagent/model, and its output is handed to the next phase:
- Identify the task — subagent: 5.6 Luna. Restate the requested outcome, inspect the relevant project files, and analyze the primary use cases, failure modes, and edge cases. Include impacts on macOS permissions, menu-bar behavior, widgets, remote control, persistence, localization, accessibility, and security when relevant.
- Write the implementation plan — subagent: 5.6 Terra. Ask the user whether they want an implementation plan written first. When accepted, load the
writing-plansskill, save the completed plan underdoc/plan/, and use it as the source of truth while implementing. For small, self-contained changes, the user may choose to proceed without a saved plan. - Load relevant skills using Context Routing — subagent: 5.6 Luna. Select only the skills whose routing conditions match the task. Do not load every skill or every skill markdown by default. See Context Routing.
- Summarize constraints — subagent: 5.6 Luna. Before changing files, briefly summarize applicable architecture rules, platform/deployment constraints, compatibility requirements, existing conventions, and any assumptions.
- Implement unit tests — subagent: 5.6 Terra. Ask whether the user wants test-driven development for the task. If yes, write or update tests before production code where practical. If no, still add or update focused regression tests when the risk warrants them. Skip this phase only when tests are not meaningful for the requested change.
- Implement the feature — subagent: 5.6 sol. Make the smallest coherent change that satisfies the task, following the applicable Context Routing guidance and the saved plan when one exists.
- Review the code — subagent: 5.6 Terra. Review the implementation and tests for correctness, architecture compliance, edge cases, concurrency safety, permissions/privacy risks, and regressions. Apply or request fixes before reporting completion. Run the narrowest relevant checks, then broader checks when appropriate, and report what changed, what was verified, and any known limitations.
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.
- 2d ago First seen · 44 lines · 1,186 tokens per session scan A 2a8a612bc504
OnlySwitch AGENTS.md is an instructions file published in the GitHub repository jacklandrin/OnlySwitch (5,902 stars, last pushed 2d ago), licensed MIT. It adds 1,186 tokens to every session, about $0.0059 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 2026-08-30.
Other instructions, from other repositories
NemoNotch CLAUDE.md
Instructions for GaoZimeng0425/NemoNotch, covering nemonotch — claude.md, project overview, tech stack, project structure and architecture.
NemoNotch AGENTS.md
Instructions for GaoZimeng0425/NemoNotch, a project described as: An interactive floating panel for the MacBook notch area, turning the notch into a multi-purpose information hub.
openscreen AGENTS.md
Instructions for getopenscreen/openscreen, covering agents.md, setup commands, development principles, project layout and code style.
codex-island CLAUDE.md
Instructions for ericjypark/codex-island, covering claude.md, release process — mandatory, the loop (3 commands), hard rules — break these and you brick auto-update and ci secrets (one-time, already configured).
skills-manager CLAUDE.md
Instructions for yibie/skills-manager, covering skills manager - development guide, tui engine: blessed, 核心原则, 参考文档 and 正确的 list 用法.
TokenEater CLAUDE.md
Instructions for AThevon/TokenEater: The full contributor and agent guide for this project lives in AGENTS.md: repo layout, architecture and data flow, the hard-won SwiftUI rules, and the build / test / release commands.