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/rome-os/rome/claude-mdgit clone --depth 1 https://github.com/rome-os/romeWrote 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/rome-os/rome/claude-md)<a href="https://agentmods.dev/instructions/rome-os/rome/claude-md"><img src="https://agentmods.dev/badge/instructions/rome-os/rome/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 | $0.01210 | $0.01210 |
| Opus 5 | $0.00605 | $0.00605 |
| Sonnet 5 | $0.00242 | $0.00242 |
| Haiku 4.5 | $0.00121 | $0.00121 |
Grade A, and why
rome 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 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 — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Rome
A pnpm monorepo. All code lives under packages/* and rome_apps/*. Runtime dependencies are declared in packages/core — the root package.json holds only the dev toolchain, and its scripts delegate to the right workspace via pnpm --filter.
packages/core/(@rome/core) — backend runtime: Hono server, agent/action/event subsystems, DB, channel adapters.src/index.tswires the object graph via dependency injection.packages/web/(rome-web) — Rsbuild SPA dashboard, served by the backend. Guardian-only.packages/ui/(@rome-os/ui) — shared React component kit for the dashboard and apps.packages/app-runtime-sdk/(@rome-os/app-runtime) andpackages/app-web-sdk/(@rome-os/app-web-sdk) — SDKs consumed by apps.packages/desktop/(rome-desktop) — Electron shell.packages/cdp-client/(rome-cdp-client) — standalone CDP client.rome_apps/*— runtime-loaded apps (plugins), each a workspace package.
Playbook
- Before opening a PR, read
docs/authoring/prs.md. - Before explaining a concept or asserting how a surface behaves, check
docs/and link instead of re-deriving. - Before building or editing an app, read
rome_apps/CLAUDE.md. - When writing prose (docs, PR text), follow
docs/authoring/WRITING.md. - When writing or editing code comments, follow
docs/authoring/comments.md. - When a PR touches
@rome-os/app-runtime,@rome-os/app-web-sdk, or@rome-os/ui, choose the Conventional Commit type by consumer impact. Test-only changes usetest:and do not bump the package; releasable changes usefeat:/fix:/feat!:(orBREAKING CHANGE:in the body). release-please publishes them to npm from Conventional Commits, with no per-PR changeset file. Full flow indocs/releases.md. - When running the fullstack local loop, use
pnpm dev:all— containerized Rome withtsx --watch, plus rome-obs, Traefik, and the Rsbuild dev server.pnpm startandpnpm start:webskip the production-shaped wiring, so reach for them only to debug the host process. - When changing a file the container does not watch (
package.json, container env), rundocker compose restart <service>. - When doing frontend-only work in
packages/web, usepnpm start:web:mock. - When verifying a change, run
pnpm typecheckon the host, thenpnpm test:unit, thenpnpm dev:alland confirm the rome container reachesRome startedindocker logs. For UI changes, also exercise the feature in the browser — type checks and tests verify code correctness, not feature correctness. - When adding a runtime dependency, declare it in
packages/core. - When bumping
@playwright/test, re-pin thelayout-invariantscontainer image in.github/workflows/ci.ymlto the matching release's digest (docker buildx imagetools inspect mcr.microsoft.com/playwright:v<version>-nobleprints it). The image ships the browsers, the pin decides the client, and they are separate edits. Check the guard step's parse ofplaywright install --dry-runin the same pass, since it reads a CLI label that upstream can reword. The guard fails readably when the image and the client disagree. - When accessing the database, go through a repository. Core schema is
packages/core/src/db/schema.ts, app schemas arerome_apps/*/db/schema.ts. - When adding config, validate it with Zod in
packages/core/src/config.tsand add it topackages/core/.env.example. - When logging, use
createLogger(component)frompackages/core/src/logger.ts.
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 Changed · +1 lines · +4 tokens per session b5af0fe5874e
- 5d ago First seen · 41 lines · 1,206 tokens per session scan A e704af8de60c
rome CLAUDE.md is an instructions file published in the GitHub repository rome-os/rome (471 stars, last pushed today), licensed MIT. It adds 1,210 tokens to every session, about $0.0060 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
hoop CLAUDE.md
Claude Code instructions for hoophq/hoop, covering claude.md, project overview, toolchain & prerequisites, architecture and module breakdown.
gini-agent AGENTS.md
AGENTS.md instructions for Open-Curiosity/gini-agent, covering gini agent instructions, shape, adrs, boundaries and branches.
zapcode CLAUDE.md
Instructions for TheUncharted/zapcode, covering claude.md, git, read agents.md first, codebase orientation and how to add a new language feature.
MathModelAgent CLAUDE.md
Claude Code instructions for jihe520/MathModelAgent, covering claude.md, 项目概述, commands, 后端 and 安装依赖.
ralph CLAUDE.md
Claude Code instructions for SantanderAI/ralph, covering ralph — instrucciones del proyecto and paridad bash ↔ powershell (importante).
memory-bridge AGENTS.md
AGENTS.md instructions for jiabaobei/memory-bridge, covering agents.md — 记忆桥项目规约(ai 协作会话必读), 版本与发布(硬性规则,所有者 2026-08-29 定), 架构铁律 and 工程习惯.