bmad-deep-recon Skill for Claude Code: Cited Research for a Decision

BMad Method skill that turns a research question into a cited report for a decision, from the web or from your own codebase.

Installs: 12.7K; on skills.sh
Stars: 54.9K; on GitHub
Files: 27; besides SKILL.md
Updated: Nov 2026; last repo commit

Ask it to research a market, a domain, a stack choice, competitors, what users say or how your own code works. It drafts a deep-research prompt, processes a report you already have, or runs the research itself. Every claim in the result needs a source.

It only works in a project where BMad is set up, because its SKILL.md runs scripts from the project's _bmad folder through uv. Run mode needs web access. The folder has 28 files including SKILL.md, with one Python helper (recon_kit.py) and its tests.

Good for

Founders and product managers researching a market or competitors
Engineers choosing between tools or stacks
Developers who want a cited answer on why their codebase works the way it does
Researchers who want a literature review with sources

Install bmad-deep-recon

Claude Code

Copy this prompt.

Install the bmad-deep-recon skill in Claude Code. Run this command:

/plugin marketplace add bmad-code-org/bmad-plugins

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

bmad-deep-recon needs BMad set up in your project: its SKILL.md runs a script from the project's _bmad folder and tells the agent to run the bmad skill's setup when that script is missing. Pick one route, not both. The marketplace manifest lists this skill's folder under the bmad-toolbox plugin, so install bmad-toolbox after you add the marketplace, plus the plugin that holds the bmad skill (the bmad-plugins README names bmad-method and bmad-toolbox). Then ask the bmad skill to run bmad setup. Skills CLI route: it asks which coding tools to install to, so choose Claude Code, then ask the bmad skill to run bmad setup. BMad needs uv for its setup and Python scripts. The README calls v7 a preview alpha with breaking changes between updates.

From the project README. Source.

What's in the skill

When it loads

Decision-grade research on a topic or this repository. Use when the user says "deep recon", "research this", "draft a research prompt", "process this research report", "market research", "domain research", "technical research", "competitor research", "what users say about X", "literature review", "help me choose between", "research the codebase", "research how X works here" or "why is X built this way". Not for a quick code question answered in chat.

Sections in SKILL.md

Overview
How you work
Resolution rules
On Activation
Research types and decision shapes
Intents
Headless Mode

Key facts

Author: BMad Code (owner of the bmad-code-org/bmad-method repo)
License: README and the plugin marketplace manifest say MIT; GitHub reports no detected license
Installs on skills.sh: 12,657 as of 2026-11-02
GitHub stars: 54,859 for the whole bmad-code-org/bmad-method repo, 2026-11-02
Last repo commit: 2026-11-01
Files: 28 files including SKILL.md (14 references, 6 research-type guides, 2 templates, 1 Python helper and its tests)
Needs: uv, BMad set up in the project, and web access for Run mode (SKILL.md)

What it includes

Writes a deep-research prompt you run in your own research tool.
Write · Draft
Files a finished report and distills a short cited summary from it.
Write · Process
Runs the research here with parallel web searches, after one plan gate.
The plan gate is the one hard stop before research starts.
Write · Run
Answers questions about this repository from its code and git history.
Runs code · Codebase
Updates or extends an existing run folder.
Write · Refresh / Deepen
Six packs: market, domain, technical, competitive, user-voice, academic-lit.
Read · Research types
Explore by default, or select to choose between candidates.
Read · Decision shape
Plans and proceeds without questions, then prints a JSON summary.
Write · Headless mode
Helper for citation checks, claim counts, staleness dates, folder names and git anchors.
Subcommands: citations, tally, staleness, slug, escape-sources, anchors.
Runs code · recon_kit.py

From the official docs

Permissions

Medium risk

It writes research files into your project, can read your repository and its git history, and searches the web in Run mode. Keep confidential topics to Draft and Process.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Run mode searches the web in parallel
SKILL.md says Run needs web access and uses parallel web fan-out. Your topic goes into those searches, so use Draft or Process for anything confidential. github.com
Codebase research reads your repository and its git history
The Codebase service answers questions from code and git history, and recon_kit.py runs git to re-check citations. It stops at a plan gate before research starts. github.com
It writes many files
Every digest, import and report section is written to a run folder as it lands. Check the folder it names before you commit your project. github.com
This is a preview release
The BMAD-METHOD README calls v7 a preview alpha with breaking changes between updates. github.com

Technical details

Scopes	Runs uv run on _bmad/scripts in your project (resolve_customization.py, resolve_config.py, memlog.py) Runs scripts/recon_kit.py, which calls git for code and history citation checks Writes a run folder with brief, imports, digests, a report and a memlog Run mode fans out parallel subagents that search the web

Questions

What does bmad-deep-recon do?

It turns a research question into a cited, decision-ready artifact. It can draft a prompt for your own research tool, process a finished report, run web research itself or research your codebase.

How do I install it?

Use the install box at the top of this page. After installing, ask the bmad skill to run bmad setup, because this skill needs BMad set up in the project.

Does it work in the Claude app or claude.ai?

This page documents no install for the Claude app. The skill needs uv, a project folder with BMad set up and, for Run mode, web access, so the box lists Claude Code, Codex and the skills CLI.

Is it safe?

It writes research files into your project and, in Run mode, sends your topic into web searches. Codebase research reads your repository and git history. Our advice: use Draft or Process for confidential topics.

Does it make things up?

Its SKILL.md says never to conclude from training data alone, and that every claim traces to a source with publisher, publication date and access date. It reports thin data as thin.

When should I not use it?

Its description says it is not for a quick code question answered in chat.

Information

updated Nov 2, 2026

Repository: bmad-code-org/bmad-method
Folder: skills/bmad-deep-recon

bmad-code-org/bmad-method skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6

README.md · github.com
README.md · github.com
marketplace.json · github.com
SKILL.md · github.com
recon_kit.py · github.com
html-briefing.md · github.com