mackit
Make a repository ready for multi-agent work with one clone and one command.
This kit gives agents a shared workspace (.agents/), a small set of
coordination rules, and a curated skill catalog inside .agents/skills/. Users do not need to design
their own handoff protocol or hunt for skills before starting: ask an agent to
install this repo, keep the default skill bundles, and start working.
It stays local-first and lightweight. There is no daemon, database, hosted service, lock server, or required dependency.
It is not trying to replace public skill marketplaces. It complements them by making a single project immediately usable by agents: install the repo-local coordination layer, seed a small curated skill set, and add external marketplace skills only after review.
Quick Start
pip install mackit
mackit adapt .
When run in a normal terminal, mackit adapt . asks whether to install all
optional curated skills, choose specific bundles, or skip optional skills. Press
Enter to accept the recommended default: install all.
Install optional curated skills?
> Yes, install all recommended skills
No, choose bundles
Skip optional skills
Select skill bundles:
[x] common
[x] workflows
[x] design
[x] research
[x] review
For local development from this repository, use PYTHONPATH=src python -m mackit.
Why Pip Install
mackit supports two distribution paths:
pip install mackitfor normal users who want the fastest path.- A GitHub checkout for agents and maintainers who need to inspect, adapt, or improve the kit.
Pip matters because the main promise of this project is low-friction adoption.
Users should not need to clone this repository, set PYTHONPATH, understand the
source layout, or copy files by hand before their project becomes agent-ready.
The installed package includes the CLI, templates, Cursor rule, .agents/
coordination files, and optional curated skills. After installation, this should work
from any repository:
mackit adapt .
GitHub remains the source of truth for review, contribution, and agent-readable implementation details. PyPI is the convenient delivery channel.
Two Use Cases
1. Start a new project
mackit new ./my-project --package-name my_project
This creates a simple Python src/ layout plus the agent-native coordination
files.
2. Adapt an existing project
mackit adapt /path/to/existing-project
The adapter is additive and idempotent:
- Existing files are not overwritten.
- Missing files are created.
AGENTS.md,CLAUDE.md, and.gitignorereceive managed sections only once.- Cursor receives
.cursor/rules/multi-agent.mdc. - Curated skills from this kit's
.agents/skills/catalog are offered during setup.
What Gets Installed
.agents/
├── README.md
├── STATE.md
├── agents/.gitkeep
├── retros/.gitkeep
├── skills/
│ ├── .gitkeep
│ ├── README.md
│ ├── registry.toml
│ └── common/
│ └── improve-this-kit/
│ └── SKILL.md
├── tasks/.gitkeep
├── improve-this-kit/
│ ├── .gitkeep
│ ├── README.md
│ ├── feedback/
│ │ ├── .gitkeep
│ │ └── README.md
│ └── proposals/
│ ├── .gitkeep
│ └── README.md
└── templates/
├── agent_state.md
├── feedback.md
├── proposal.md
├── retro.md
├── skill.md
└── task_state.md
.cursor/rules/multi-agent.mdc
AGENTS.md
CLAUDE.md
.gitignore
src/<package_name>/ # only for `new`, or when --package-name is passed
Selected optional curated skills are copied from this kit's .agents/skills/ into the
target .agents/skills/, preserving grouped paths such as
design/image-to-editable-slides/.
Design Principles
AGENTS.mdis the cross-tool canonical contract.CLAUDE.mdis a thin Claude Code shim..cursor/rules/is a Cursor-native adapter..agents/stores runtime coordination, not long-term design knowledge..agents/retros/stores raw post-task learning drafts..agents/skills/stores reviewed, reusable skills for future agents..agents/improve-this-kit/feedback/stores raw usage feedback about the kit..agents/improve-this-kit/proposals/stores review-gated kit improvement proposals and is trackable by default.improve-this-kitturns real usage feedback into reviewed proposals and tests..agents/skills/registry.tomldescribes curated skills and bundles..agents/skills/external/README.mdtells agents how to recommend external skills safely.- Design decisions should move to your real docs, ADRs, or README.
- Work isolation should still prefer branches or worktrees for heavy parallelism.
Self-Improvement Loop
The kit is designed to improve from real usage without letting agents rewrite it from a single anecdote:
- Agents write raw usage notes in
.agents/improve-this-kit/feedback/. - Repeatable issues become proposals in
.agents/improve-this-kit/proposals/. - Proposals require explicit review/approval before implementation.
- Accepted changes must include or update an idempotency check, fixture, or test.
- Reviewed, reusable know-how can become a skill in
.agents/skills/.
Use .agents/skills/common/improve-this-kit/SKILL.md for the review-gated workflow.
Curated Skill Catalog
The repo ships its maintained catalog directly in .agents/skills/. During
manual terminal use, mackit asks whether to install all optional curated
skills, choose specific bundles, or skip them:
mackit adapt /path/to/project
For automation and agents, use explicit flags so the command never waits for input:
mackit adapt /path/to/project --yes
mackit adapt /path/to/project --skill-bundles common,design
mackit adapt /path/to/project --no-curated-skills
Use list-skill-bundles to inspect available bundles.
External skill discovery is optional. Agents may recommend external skills, but
must report candidates and wait for user approval before copying them. See
.agents/skills/external/README.md.
For Agents
If a user gives you this repository link and asks you to make their project
agent-ready, read INSTALL_FOR_AGENTS.md first.
For public sharing, see docs/external_download.md.
Release files for mackit 0.1.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| mackit-0.1.2.tar.gz | 18.8 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| mackit-0.1.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 41.8 kB
Release files / mackit-0.1.2.tar.gz
| Download URL | mackit-0.1.2.tar.gz |
|---|---|
| Size | 18.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
0b19e674c7010a930df6e8ff6ea3a82bdfdfae2afd50f7a022680b43b9dd3129
|
|
BLAKE2b-256 checksum How to use checksums |
f957bb577cd10d243d4d9c59a8e6ab99826a3cbb68516035a02c057e5d2f4a0c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.12
|
Release files / mackit-0.1.2-py3-none-any.whl
| Download URL | mackit-0.1.2-py3-none-any.whl |
|---|---|
| Size | 23.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
dbd260392897b65dc6c931c37d5977f92661e3c8fd2fc32a4e277a503e56d7b3
|
|
BLAKE2b-256 checksum How to use checksums |
b0288ac1089d0003b40b93805992ee5d9f98271e05c576f330cec8f9005f3851
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.12
|