ai-research-arm: Command for Claude Code

.claude/commands/gen-research.md

gen-research is a command for Claude Code from guzus/ai-research-arm. It costs 51 tokens per session (4,896 once invoked), scanned A, original, MIT.

A research command that investigates a topic using original sources and writes the result as an HTML article. It first asks whether to run in the current chat or through a GitHub Actions workflow, which runs automated jobs in GitHub.

In plain words
What is it for?
It helps research a supplied topic, verify information, and save a finished article in the repository.
Why use it?
It provides a choice between a quicker local run and a longer remote run with deeper research, before starting the work.

Command for Claude Code

Written for Claude Code: argument-hint in frontmatter. Also seen: mentions subagents; names the AskUserQuestion tool.

This is guzus/ai-research-arm's own configuration. It tells Claude Code how to work on ai-research-arm itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything ai-research-arm configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is uv run python scripts/check_generative_research.py /tmp/gen-research.ara.md \.

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/commands/gen-research.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.

agentmods badge for gen-research

README.md
[![agentmods](https://agentmods.dev/badge/commands/guzus/ai-research-arm/gen-research/github.svg)](https://agentmods.dev/commands/guzus/ai-research-arm/gen-research)
Your own site
<a href="https://agentmods.dev/commands/guzus/ai-research-arm/gen-research"><img src="https://agentmods.dev/badge/commands/guzus/ai-research-arm/gen-research/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.

agentmods 80×15 button for gen-research

Your own site · 80×15
<a href="https://agentmods.dev/commands/guzus/ai-research-arm/gen-research"><img src="https://agentmods.dev/badge/commands/guzus/ai-research-arm/gen-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 51 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,896 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00051 $0.04896
Opus 5 $0.00026 $0.02448
Sonnet 5 $0.00010 $0.00979
Haiku 4.5 $0.00005 $0.00490

Measured 10d ago against content hash 0595240ee5c4, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

gen-research scanned grade A with 1 finding 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 10d 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.

Makes network callslowCapability

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

- `curl -sL <url> -o /tmp/x.pdf && pdftotext /tmp/x.pdf - | head -c 60000`
.claude/commands/gen-research.md · 441 lines

How it starts

The opening of the file, as written. The whole thing — 441 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Generative Research

User-supplied topic: $ARGUMENTS

If the topic above is empty or just whitespace, ask the user for a topic and stop until they answer.

Step 1 — choose execution mode (always ask)

Use AskUserQuestion to ask the user how they want this run executed. DO NOT skip this prompt — the two modes have very different cost, latency, and depth profiles. Present these two options:

  • Run locally (in this chat) — ~5–15 min. Deep pipeline runs in the current session using sub-agents (Task tool), evidence packets, and a verifier pass. Article lands as a local commit in research/generative/ (not pushed). Recommended for topics you want to iterate on, or when you want to read the result immediately.
  • Route to GitHub workflow — ~20–45 min on the self-hosted runner. Dispatches generative-research.yml with backend=deepseek-v4-flash by default. Your chat is freed instantly. The runner produces a longer article (4500–7000 words target, ≥20 cited references) and pushes the commit directly to main. Recommended for substantive topics where you want maximum depth without spending in-chat time.

After the user picks:

  • Route → follow PATH A below.
  • Run locally → follow PATH B below.

PATH A — Route to the GitHub workflow

Dispatch the workflow with the user's topic. Use deepseek-v4-flash as the backend unless the user explicitly asked for Claude in their brief.

gh workflow run generative-research.yml \
  -f topic="<the exact topic the user gave>" \
  -f backend=deepseek-v4-flash

Then wait briefly for the run to register and surface its URL:

sleep 5
gh run list --workflow=generative-research.yml --limit 1 \
  --json databaseId,status,url,createdAt,event \
  --jq '.[0]'

Report to the user in one short message:

  • "Dispatched to GitHub workflow."
  • The run URL.
  • The command to watch live: gh run watch <id>.
  • Expected runtime (~20–45 min) and the destination (research/generative/). The runner pushes the commit, so the article shows up at https://ara.guzus.xyz/research/ after the next Vercel deploy.

Read the full file on GitHub · 441 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.

  1. 10d ago First seen · 441 lines · 51 tokens per session scan A 0595240ee5c4

Subscribe to this mod's changes

gen-research is a command published in the GitHub repository guzus/ai-research-arm (11 stars, last pushed yesterday), licensed MIT. It adds 51 tokens to every session and 4,896 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.