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/iishyfishyy/operator-oss/claude-mdgit clone --depth 1 https://github.com/iishyfishyy/operator-ossWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/iishyfishyy/operator-oss/claude-md)<a href="https://agentmods.dev/instructions/iishyfishyy/operator-oss/claude-md"><img src="https://agentmods.dev/badge/instructions/iishyfishyy/operator-oss/claude-md.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.03595 | $0.03595 |
| Opus 5 | $0.01798 | $0.01798 |
| Sonnet 5 | $0.00719 | $0.00719 |
| Haiku 4.5 | $0.00360 | $0.00360 |
Grade A, and why
operator-oss CLAUDE.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 5d 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 — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Orchestrator — a local-first web app that runs many Claude Code sessions in parallel across multiple projects from one screen. Each project carries reusable context + a working directory; each task is its own Claude Code session in its own git worktree, driven by @anthropic-ai/claude-agent-sdk against the user's local Claude login (no API key). (A hosted version, getoperator.dev, lives in a separate private repo that overlays this one — see "Repo split" below.)
Commands
npm run dev— app (:3000,server.js) + pty sidecar (:3001,pty-server.js) via concurrently.npm run dev:next/npm run ptyrun them separately.npm run build(turbopack) thennpm startfor production.npm test— vitest, serial on purpose (tests spawn many real git subprocesses). Single file:npx vitest run tests/merge.test.ts.npm run test:e2e— Playwright end-to-end suite: builds, then boots the real prod server against a hermetic temp instance with the deterministic mock agent (lib/agents/mock/, registered only whenORCH_E2E_MOCK_AGENT=1) and drives onboarding → project → task → turn → diff → merge through the UI.npm run preflight= unit + e2e, the pre-push gate. Seee2e/README.md(mock-turn directives, selector conventions, staleness gotcha: the server runs the built bundle).- No lint script; TypeScript is strict, path alias
@/*→ repo root (mirrored invitest.config.ts).
Architecture
Three processes/entrypoints, one origin:
server.js— custom Next.js server (plain Node, CommonJS). Fronts Next on one port, proxies/ptyWebSocket upgrades to the sidecar, forwards dev HMR upgrades to Next, enforces origin auth on WebSocket upgrades (middleware never sees upgrades — this file is the auth boundary for the terminal), and dispatches public service hostnames (<slug>--<appHost>) throughlib/service-router.mjs.pty-server.js— node-pty sidecar, bound to127.0.0.1only; never exposed directly.- Next app — UI in
app/, REST underapp/api/, server logic inlib/.
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.
- 5d ago First seen · 78 lines · 3,595 tokens per session scan A e7da8d9fef43
operator-oss CLAUDE.md is an instructions file published in the GitHub repository iishyfishyy/operator-oss (201 stars, last pushed 15d ago), licensed Apache-2.0. It adds 3,595 tokens to every session, about $0.0180 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
git-worktree-runner copilot-instructions.md
Instructions for coderabbitai/git-worktree-runner, covering copilot instructions, architecture & flow, key concepts, adapter contract and manual testing (essential subset).
git-worktree-runner completions.instructions.md
Instructions for coderabbitai/git-worktree-runner, covering completions instructions, when to update completions, file responsibilities, implementation pattern and testing completions.
git-worktree-runner testing.instructions.md
Instructions for coderabbitai/git-worktree-runner, covering testing instructions, basic create/remove, branch sanitization, remote branch (if exists) and local existing branch.
git-worktree-runner ai.instructions.md
Instructions for coderabbitai/git-worktree-runner, covering ai adapter instructions, when to use a file vs registry, adding a standard ai tool (registry), adding a custom ai tool (file override) and toolname adapter.
git-worktree-runner editor.instructions.md
Instructions for coderabbitai/git-worktree-runner, covering editor adapter instructions, when to use a file vs registry, adding a standard editor (registry), adding a custom editor (file override) and editorname adapter.
mulmoterminal CLAUDE.md
Claude Code instructions for receptron/mulmoterminal, covering claude.md — mulmoterminal, stack & package manager, run after changes, import a component at module scope, never inside a test and no emojis.