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/emaitchess/refd/claude-mdgit clone --depth 1 https://github.com/emaitchess/refdWhat 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 | $0.06196 | $0.06196 |
| Opus 5 | $0.03098 | $0.03098 |
| Sonnet 5 | $0.01239 | $0.01239 |
| Haiku 4.5 | $0.00620 | $0.00620 |
Grade A, and why
refd 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 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.
This is a copy
97% identical to refd AGENTS.md — 8 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
AGENTS.md is the tool-neutral twin of this file — apply any change to both.
What this is
refd (refd.ai) — open-source AI search monitoring. Each workspace tracks one brand's visibility, mentions, citations, and rank across AI answer surfaces: ChatGPT, Perplexity, Gemini, Google AI Mode (BrightData dataset scrapers), and Google AI Overviews (BrightData SERP API). BrightData is the only data provider. Available hosted (refd.ai) or self-hosted; born tracking mrmr, since generalized. Public copy never pairs "refd" with "aeo".
Reference docs: docs/DESIGN.md (design system — all UI work follows it), docs/METRICS.md (scoring and aggregation contract), and docs/mcp.md (remote MCP and OAuth usage).
Monorepo layout
Bun workspaces (apps/*, packages/*) — three independently deployed Cloudflare Workers plus a shared package (docs/plan-three-worker.md).
apps/api(@refd/api,api.refd.ai) — the Worker: Hono API, OAuth, MCP, cron trigger, queue consumer, every binding (D1, R2, KV, Queues, AI, Browser, rate limits), and secrets (apps/api/.dev.varslocally). The only Worker that declares bindings, cron, or queue consumers — CI enforces this viascripts/check-worker-topology.sh.apps/dashboard(@refd/dashboard,dash.refd.ai) — the React + Vite SPA as an assets-only Worker (@/*→src/*).noindex, with its ownpublic/robots.txt+public/_headers(CSP allowsapi.refd.ai) and no landing route. It calls the API cross-origin, so its client builds absolute URLs fromVITE_API_ORIGINand sendscredentials: 'include'.apps/web(@refd/web,refd.ai) — the public site: a static Astro build fronted by a tiny Worker (src/worker.ts) that does homepage Markdown negotiation over the assets. Owns the indexable SEO assets (apps/web/public/),@astrojs/sitemap, JSON-LD, and the static 404. Pure.astro(Biome-excluded;astro checkcan't run on the repo's TS 7, soastro buildvalidates the templates); CTAs link todash.refd.ai.packages/core(@refd/core) — runtime-neutral code shared across Workers (config,surfaces,mentions,workspaces,site-metadata,blocks,metric-copy,glossary,public-pages), imported via subpaths like@refd/core/config. No build step; consumers bundle the TS source.
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 · 95 lines · 6,196 tokens per session scan A 8b91ab6e865d
refd CLAUDE.md is an instructions file published in the GitHub repository emaitchess/refd (0 stars, last pushed 1mo ago), licensed MIT. It adds 6,196 tokens to every session, about $0.0310 per session on Opus 5. A static security scan graded it A with 0 findings. It is 97% identical to refd AGENTS.md, differing in 8 lines, and is treated as a copy.
Other instructions, from other repositories
aeo-radar CLAUDE.md
Instructions for hellowalt/aeo-radar, covering claude.md — dev playbook for aeo radar, tech stack, common commands, crawl (defaults to headless) and seeding & analysis.
geoskills CLAUDE.md
Instructions for Cognitic-Labs/geoskills, covering geoskills — ai agent guide, repository overview, directory structure, skill internal structure and skill.md conventions.
agents AGENTS.md
AGENTS.md instructions for unifapi-agent/agents, covering agents, role agents, seo agent, ai visibility agent and local seo agent.
potato CLAUDE.md
Instructions for onism1767-creator/potato, covering claude.md — 给 claude code 的厂规（本仓库的最高行为约束）, 0. 一句话项目定位（先理解再动手）, 🟥 叙事红线（写注释、文案、报告、readme 时都适用）, 1. 架构层级与不可跨越的边界（🟥 核心红线） and 🟥 边界红线.
loamwright-SEO-Skill AGENTS.md
Instructions for XuanRanL/loamwright-SEO-Skill, covering agents.md — universal host intent router, intent → skill routing, project initialization, article creation (most common) and single-phase entry.
loamwright-SEO-Skill CLAUDE.md
Instructions for XuanRanL/loamwright-SEO-Skill, covering project: xuanran seo blog writer (claude code plugin), what this is, rule 4 — always verify the live url after publish, where these rules are enforced and plugin install sync.