bkit-gemini plugin
A Gemini CLI extension on the PDCA cycle: 21 agents, 35 skills and 24 commands for planning, design, build and analysis
- 66 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Agents read and edit project files and act without confirmation at high automation levels
- The extension writes documents into `docs/` and keeps agent memory across sessions
Install
- Works in
Manual install
gemini extensions install https://github.com/ww-w-ai/bkit-geminiAfter installing, run `gemini -p "bkit Hi"`: the extension reports its activation.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agents read and edit project files and act without confirmation at high automation levels
- The extension writes documents into `docs/` and keeps agent memory across sessions
This is third-party code. Review the repository files before installing.
What it does
bkit-gemini is a Gemini CLI extension that drives development through the PDCA cycle: plan, design, do, check. The `/pdca plan`, `design`, `implement`, `analyze`, `iterate` and `report` commands write documents into `docs/01-plan`, `docs/02-design`, `docs/03-analysis` and `docs/04-report`, so every feature leaves a trail of decisions. The kit ships 21 specialized agents, 35 skills and 24 TOML commands. You pick a project level with `/starter` for a static site, `/dynamic` for a full stack or `/enterprise` for microservices. The request language is detected automatically across eight languages, and `/control level <N>` sets automation from L0 (manual) to L4 (full auto with safety gates). Agents keep their own memory across sessions. Context is built in three layers: layered, just-in-time and memory. This is the Gemini CLI port of bkit for Claude Code: the methodology is shared, while hooks and policies are adapted to Gemini CLI.
Who it is for. Developers on Gemini CLI who want one process from plan to report and documents that stay in the repository.
Good fit when
- A new feature has to go step by step: plan, design and implementation with documents in `docs/`
- A project starts from scratch and needs the `/starter`, `/dynamic` or `/enterprise` level
- The implementation has to be checked against the design: the gap-detector agent finds gaps and `pdca-iterator` repeats the cycle
- Agent autonomy has to change with `/control level`, from manual to full automation
Not a fit when
- You work in Claude Code: there is a separate bkit edition for it
- The machine runs a Gemini CLI older than 0.34.0 or has no Node.js 18
- The task is small and five documents per feature would only slow you down
Example request
Run /pdca plan login and draft a plan for the login feature with a document in docs/01-planLimitations
- It needs Gemini CLI 0.34.0 or newer (tested through 0.42.0) and Node.js 18 or newer; the extension is plain JavaScript with no native bindings.
- Reference docs live in `docs/reference/`, and the README keeps a short summary with links. Russian is not among the eight auto-detected languages.
- Automation levels L3 and L4 hand more decisions to agents without confirmation, so start at L0 or L1.
- It runs on macOS, Linux, WSL and Git Bash.
How to disable. Remove the extension with `gemini extensions uninstall bkit`; documents created in `docs/` stay in your project.
FAQ
How do I install bkit for Gemini CLI?
Run `gemini extensions install https://github.com/ww-w-ai/bkit-gemini`, then `gemini -p "bkit Hi"`. The reply says bkit is activated at the Starter level.
What is PDCA in bkit?
A four-step cycle: plan, design, do and check. Each `/pdca` command writes a document into its own folder under `docs/`.
How do `/starter`, `/dynamic` and `/enterprise` differ?
They are project levels: `/starter` for static HTML and CSS sites, `/dynamic` for a full stack and `/enterprise` for AWS and Kubernetes microservices.
How does bkit-gemini differ from bkit for Claude Code?
The PDCA method and context layers are shared, while hooks and policies are adapted to the Gemini CLI contract. The mapping between editions is in `docs/reference/architecture.md`.
How do I limit agent autonomy?
Use `/control level <N>`: L0 leaves every decision to you and L4 turns on full auto with safety gates.
Discussion
No questions or reviews yet.
Related
PluginsA library of 89 subagents for Gemini CLI and Antigravity CLI: review, security, testing, DevOps, data and language experts.
An MCP server that lets the agent call Gemini through its CLI to analyze large files and whole codebases
Gem Team installs 12 agent roles and a plan, TDD, verify, learn process in place of one-off code generation
An OpenAI Codex CLI extension: 27 skills, 16 MCP tools and AGENTS.md rules for PDCA development with a mandatory design document
Ten Claude Code commands that run Codex and Gemini as helpers: Claude plans, Codex writes code, Gemini reviews and reads mockups
A console chat with ChatGPT, Claude, Gemini and Cohere: custom assistants in YAML, one-shot prompts from scripts and shell commands you edit before running
README in short
The README briefly presents bkit as a Gemini CLI extension, lists capabilities and compatibility and points to QUICKSTART and the references in `docs/reference/`.