Think Better CLI
Methods for an agent in one CLI: comparing options, solving problems and changing code in steps, 20 slash commands in all.
- 125 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The CLI writes skills and commands into project folders, and the skills run local Python scripts that read project files and git history.
Install
- Works in
Manual install
brew tap HoangTheQuyen/think-better https://github.com/HoangTheQuyen/think-better && brew install think-betterThe first README option, for macOS and Linux. Then in the project: `think-better init --ai claude`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The CLI writes skills and commands into project folders, and the skills run local Python scripts that read project files and git history.
This is third-party code. Review the repository files before installing.
What it does
Think Better installs three skills and a set of slash commands into a project. `/decide` compares options against weighted criteria and flags cognitive biases. `/solve` takes a problem through seven steps, from a precise definition to a summary for leadership. `/code` changes code in seven steps too, and each step needs evidence: a failing test before the fix, then test, lint and build output after it. The skills run locally on Python 3 with the standard library only. Search covers a knowledge base of 326 items: decision frameworks, decomposition techniques and 44 common error messages. `/code` reads the project: it maps stack-trace frames to files, finds where symbols are defined, checks recent commits and picks the test commands. Asking to "save step-by-step" writes a folder of Markdown files, and `/code.resume` continues at the first unmet step.
Who it is for. Developers and team leads who want an agent to follow a method instead of a first guess: stack choice, root-cause hunting, refactoring.
Good fit when
- A team is choosing between Postgres, MongoDB and DynamoDB and needs a weighted criteria matrix.
- Checkout throws a `TypeError` after yesterday's release, and the cause is traced step by step with a failing test.
- A metric dropped after a pricing change and the problem is split into a hypothesis tree.
- A refactor follows a plan: a change map, then the project checks after each step.
- A decision is recorded in a journal to compare with the real outcome a month later.
Not a fit when
- The question is simple and a quick chat answer needs no seven steps.
- Requests are written in Russian: the knowledge base and trigger phrases are English, so skill matching is weaker.
- The environment has no Python 3, which the skill scripts need.
- The agent is not Claude Code, GitHub Copilot, Antigravity or OpenCode.
Example request
/decide Postgres vs MongoDB vs DynamoDB for our order service, criteria and weights first, scores afterLimitations
Python 3 is required (check with `think-better check`). The knowledge base and trigger phrases are in English, so write requests in English for the best match. `init` supports four tools: Claude Code, GitHub Copilot, Antigravity and OpenCode. For Copilot a global install adds the skills but not the slash commands. After upgrading the binary run `think-better update`; files you edited are kept. The install script downloads a binary and verifies its SHA-256 checksum, but read it before piping `curl` into `sh`.
How to disable. Delete the skill and slash-command folders from `.claude/skills/` and `.claude/commands/` (see the install table for other tools), then remove the `think-better` binary.
FAQ
How do I install Think Better for Claude Code?
Install the binary (Homebrew, Scoop, the install script or `go install`), then run `think-better init --ai claude` in the project. Skills land in `.claude/skills/` and slash commands in `.claude/commands/`.
How do `/decide`, `/solve` and `/code` differ?
`/decide` chooses between options, `/solve` takes apart a business or product problem, `/code` changes code. Each has depth variants: `.quick`, the plain command, `.deep` and `.exec` for a leadership summary.
Can I install a single skill?
Yes. The `--skill code-solving` flag installs one skill with its commands, and `--exclude-command code.perf` leaves out a single command.
What should I do after upgrading the binary?
Run `think-better update`. It refreshes every install and keeps files you edited yourself.
Is Python required?
Yes, Python 3 with the standard library: the skills run its scripts. Check the environment with `think-better check`.
Discussion
No questions or reviews yet.
Related
CLI toolsThe /improve skill: a strong model audits a repository and writes self-contained fix plans, and you hand the execution to a cheaper model
A cleanup skill for agent-written code: finds surplus tests, self-checks and fallbacks and deletes them, keeping behavior
kyy-logs code-review-skill
code-review-skill
A code review skill with four independent probes, evidence-backed findings and a merge gate derived from JSON
A set of 116 skills, 44 agents and safety hooks for Claude Code, Cursor, Codex, Gemini CLI and nine more editors
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
README in short
The project rests on the idea that an agent needs a method, not only a prompt. The method comes from fixed steps, evidence gates and a local knowledge base; the `examples/` folder shows full walk-throughs.