research

A command for researching claims that need outside evidence and saving the findings in the project's source or context notes.

In plain words
What is it for?
Use it to fact-check a source, investigate a focused question, compare evidence, and record conclusions for later work.
Why use it?
It helps verify specific claims without turning the task into a broad, unfocused summary. It also keeps new evidence linked to the relevant research material.

Command for Claude Code

Install

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.

agentmods
npx agentmods add commands/0xchamin/brain-kit/research
Clone the repo
git clone --depth 1 https://github.com/0xchamin/brain-kit

Made for: Claude Code.

Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 853 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00018 $0.00853
Opus 5 $0.00009 $0.00426
Sonnet 5 $0.00004 $0.00171
Haiku 4.5 $0.00002 $0.00085

Measured 2d ago against content hash bc026d9b36b5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

research 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.

.claude/commands/research.md · 60 lines

How it starts

The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/research - deep research (external evidence)

The contract is AGENTS.md § "Deep research on request". Read it now and follow it - this file is only the Claude Code wrapper for that stage. Do not duplicate the contract here; if the two ever disagree, AGENTS.md wins.

Adopt fact-checker + synthesizer (+ mentor if the goal is teaching the concept).

Target

$ARGUMENTS

If that names a source folder, research that source's gated nodes. If it names a topic or a bare question, research that - but still land the output in a context/ note under the most relevant source, or in reports/ if it spans several.

Run it

  1. Orient. Read the source's SOURCE.md and nodes.md. Pick the research targets by ID: single-leg nodes, anything needs-check, recorded divergences, and the LEARNING.md open questions. Target claims, not the subject - open-ended topic research makes you a summarizer.
  2. Read the brain before the web. grep the root INDEX.md, brain/topics/*.md and brain/claims.md. A prior source may already answer this; that link beats a fresh fetch.
  3. Search and fetch, favouring T1-T3 (see the tier table in AGENTS.md). Deliberately attempt the cross-domain hop - the established name for this idea in an older discipline. Respect the budget: ≤ 8 searches, ≤ 12 fetches; stop early on two independent agreeing sources, or when a pass surfaces nothing new.
  4. Apply the independence rule. A companion repo, a vendor blog restating that vendor's own talk, or same-lab work is the same leg wearing a different hat: record it, cite it, but do not raise confidence.
  5. Write sources/<id>/context/<NN>_<slug>.md using the structure in sources/_TEMPLATE/context/README.md. Every finding gets a verdict (supports / contradicts / refines / no-evidence), a tier, and an independence call. End with a Confidence assessment - never interrupt with clarifying questions; state assumptions there instead.
  6. Feed it back in the same pass: update node confidences in nodes.md (pointing at the context note), cite external support in brain/claims.md, add terms to brain/glossary.md, set SOURCE.md Status, and append to brain/log.md. Keep external findings out of LEARNING.md's body - it may cite the note, never absorb it.
  7. Report back: one paragraph on what changed - which claims got stronger, which got weaker, which found nothing - then git diff.

Read the full file on GitHub · 60 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. 2d ago First seen · 60 lines · 18 tokens per session scan A bc026d9b36b5

Subscribe to this mod's changes

research is a command published in the GitHub repository 0xchamin/brain-kit (2 stars, last pushed 1mo ago), licensed MIT. It adds 18 tokens to every session and 853 once invoked, about $0.0001 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 2026-08-31.