What is ai-research-arm CLAUDE.md?

Project instructions for AI Research Arm, an automated AI-news pipeline that gathers information from sources such as Twitter/X, RSS feeds, Bluesky, Hacker News, Reddit, and arXiv, then publishes research and digests to a web app.

They guide agents working on source collection, scheduled aggregation, research synthesis, model timelines, articles, workflows, and the dashboard.

They explain the project’s architecture, operating assumptions, and which parts require accounts or special infrastructure.

How guzus/ai-research-arm configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to guzus/ai-research-arm. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/guzus/ai-research-arm/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/guzus/ai-research-arm

Made for: Claude Code.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/guzus/ai-research-arm/claude-md/github.svg)](https://agentmods.dev/instructions/guzus/ai-research-arm/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/guzus/ai-research-arm/claude-md"><img src="https://agentmods.dev/badge/instructions/guzus/ai-research-arm/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/guzus/ai-research-arm/claude-md"><img src="https://agentmods.dev/badge/instructions/guzus/ai-research-arm/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 22,936 This file is loaded in full into every session.

When invoked 22,936 The same file — it is already loaded in full.

Security scan B 2 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

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 5.1	$0.22936	$0.22936
Opus 5.5	$0.09174	$0.09174
Sonnet 5.5	$0.04587	$0.04587
Haiku 4.5	$0.02294	$0.02294

Measured today against content hash a61ae190bcde, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade B, and why

ai-research-arm CLAUDE.md scanned grade B with 2 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 today.

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.

Reads MCP configurationmediumAgent snooping

mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.

| **Claude Sonnet 5.5 (isolated Claude Code)** | Standby for strict isolated routes (served them 2026-10-31 to 2026-11-05; currently quota-blocked); `claude-isolated-canary.yml` | `CLAUDE_CODE_OAUTH_TOKEN` | Claude Code

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

| `market-quotes.yml` | every 3h `:53` | Deterministic refresh of `research/market/quotes.json` from Yahoo for every wiki entity carrying `market` frontmatter — the price row in the wiki hover card and on the entity page

CLAUDE.md · 940 lines

How it starts

The opening of the file, as written. The whole thing — 940 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Operating guide for AI agents working in this repo. Read this before touching workflows, research output, or the dashboard.

New here, or just found this on GitHub? This file is the operator's manual for the live deployment and assumes the maintainer's infrastructure (self-hosted runners, S3, telemetry, X cookies). If you are exploring or forking the project, start with README.md: its Quickstart runs with no accounts, and the License and "What you can run vs. what needs accounts" sections explain what is reproducible outside the original deployment.

Project Overview

ai-research-arm (ARA) is an automated AI-news intelligence pipeline. GitHub Actions on the self-hosted gunux Linux runner aggregate signal from Twitter/X, RSS, Bluesky, Hacker News, Reddit, and arXiv; synthesize it into daily digests, model-release timelines, and long-form generative articles; and publish everything to a Vite/TS SPA at ara.guzus.xyz.

Core Pipeline Architecture

sources (Twitter/RSS/Bluesky/HN/Reddit/arXiv, SEC filings, podcasts + YouTube via tuber)
    → aggregator workflows (hourly/4h/daily) write into research/<source>/
    → synthesis workflows (digest, news research, model timeline) read research/
    → output workflows (front-page, generative-research) produce shareable artifacts
    → aradio.yml (as input lanes finish) CRUDs a situation board and airs two-host audio bulletins
    → dashboard prebuild copies research/* into dashboard/public/research/
    → Railway rebuilds the root Dockerfile image and deploys on every push to main

Daily-improve runs at the tail of the cycle, reads yesterday's output, and opens a PR with methodology fixes.

Key Components

Path	What it is
`ARA_DSL.md`	Source format for generative-research articles; `scripts/compile_ara.py` turns `.ara.md` into a validated `<article>` fragment.
`ARA_CATALOG.json` + `COMPONENTS.md`	The `ara-*` class allowlist the validator loads, and its human reference. Kept in lockstep by CI — see "Component catalog" below.
`data/agent-backends.json`	Routing SSOT: per-lane backend, profile table, ordered fallback chain. See Backends.
`data/artifact-slos.json`	Artifact/SLO SSOT: producer, exact output paths, cadence kind, freshness threshold, validators, content policy, and degraded behavior. Event/on-demand artifacts are inventoried without false clock-based paging.
`docs/*.md`	Per-lane contracts, read at runtime by the agents and enforced by the matching validator: `model-tickets`, `model-onboarding`, `wiki-schema`, `okf`, `headline-dedupe`, `blog-subscriptions`, `twitter-account-curation`, `twitter-model-ab`, `generative-research-backends`, `backend-matrix` (generated), `claim-store`, `hooker-telemetry`, `aradio` (the Aradio contract: gate, inbox, editor ops, validator, writer, published files). `docs/archive/` holds superseded docs + improvement logs.
`dashboard/`	Vite + Bun + TypeScript SPA. `prebuild.mjs` copies `research/*` into `public/research/` and emits `manifest.json`.
`Dockerfile` + `Caddyfile` + `railway.json`	The Railway stack serving ara.guzus.xyz behind Cloudflare (responses carry `x-railway-edge`): bun build → Caddy serves `dashboard/dist`; `railway.json` pins the DOCKERFILE builder + `/` healthcheck. Vercel no longer serves the domain — rule 3.
`data/`, `research/`	Static lookup data for aggregation scripts; all generated artifacts (subdir map in "Output Locations").

Read the full file on GitHub · 940 lines

Changes

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.

today Changed · +7 lines · +101 tokens per session a61ae190bcde
yesterday Changed · -10 lines · -298 tokens per session 9a4368248d54
4d ago Changed · +1 tokens per session 0812a7bc6566
5d ago Changed · +25 lines · +2,192 tokens per session scan A → B 991c651c83ad
6d ago Changed · +64 lines · +2,521 tokens per session 3ea1ab1d197e
12d ago Changed · +27 lines · +330 tokens per session 10da094d7364
21d ago Changed 092e81c2eeb0
1mo ago Changed · +16 lines · +281 tokens per session 802e1a90cadc
2mo ago Changed · +10 lines · +202 tokens per session 9b04d1c5542a
2mo ago First seen · 801 lines · 17,606 tokens per session scan A 177b9115d844

Subscribe to this mod's changes

ai-research-arm CLAUDE.md is an instructions file published in the GitHub repository guzus/ai-research-arm (14 stars, last pushed today), licensed MIT. It adds 22,936 tokens to every session, about $0.0917 per session on Opus 5.5. A static security scan graded it B with 2 findings (reads mcp configuration, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

ai ai-agents claude github-actions news

ai-research-arm CLAUDE.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
ai-research-arm CLAUDE.mdguzus/ai-research-arm	Instructions file	14	today	MIT	B	22,936	Claude Code
vscode buildNext.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	9,719	GitHub Copilot
next.js AGENTS.mdvercel/next.js	Instructions file	143k	yesterday	MIT	A	7,461	Codex, OpenCode
langchain AGENTS.mdlangchain-ai/langchain	Instructions file	149k	yesterday	MIT	A	4,680	Codex, OpenCode
vscode oss-third-party-notices.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	5,516	GitHub Copilot
vscode copilot-instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	3,093	GitHub Copilot
deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness	Instructions file	274k	2d ago	MIT	A	4,313	Codex, OpenCode

ai-research-arm: Instructions file for Claude Code