claude-config skill
A Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
- 914 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skills run Python scripts and sub-agents that edit project files
- Merging into an existing `~/.claude` may overwrite personal settings on conflicts
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/claude-configDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills run Python scripts and sub-agents that edit project files
- Merging into an existing `~/.claude` may overwrite personal settings on conflicts
This is third-party code. Review the repository files before installing.
What it does
claude-config splits work on a non-trivial change into three steps: exploration, planning and execution. The `planner` skill writes the plan to a file together with the decisions made and the options rejected, and at execution it delegates milestones to sub-agents and writes no code itself. The `architect`, `developer`, `debugger`, `technical-writer` and `quality-reviewer` agents sit in the `agents` folder. The next milestone starts only after the previous one passes review. For context the repository introduces a two-file scheme in every folder. `CLAUDE.md` stays a short index with hints on when to open a file, and `README.md` holds non-obvious architecture knowledge. The technical writer agent enforces budgets of roughly 200 tokens for the first and 500 for the second. Separate skills cover narrow tasks: `deepthink` and `problem-analysis` for reasoning, `decision-critic` for stress-testing a decision, `codebase-analysis`, `refactor` with a prioritized list of problems, `doc-sync`, `prompt-engineer`, `arxiv-to-md` and `cc-history`.
Who it is for. Backend developers and maintainers of large codebases who drive non-trivial changes in Claude Code and want the agent to plan and pass review before merging.
Good fit when
- A task touches many files and before code you need a plan with recorded decisions that survives a context clear
- You must choose between several approaches, for example artifact versioning in S3 or a logging library
- LLM code works but has duplicated functions and growing huge methods, and you need a prioritized problem list
- Repository documentation drifted apart after a refactor and has to be put in order
- A sub-agent prompt misbehaves and has to be optimized with the edits explained
Not a fit when
- A one-line fix: the plan and review cycle burns time for nothing
- There is no Python 3: the skills launch their scripts through `python3`
- A front-end project: the author admits the workflow was tuned on backend work and has not been tested for interfaces
Example request
Use the planner to write a plan to plans/logging.md for moving from the homegrown Log() to a logging library with file rotationLimitations
The configuration is opinionated, the author describes it as a personal approach and ran no formal tests. The skills run Python scripts with a command like `python3 -m skills.deepthink.think --step 1` from the `.claude/skills/scripts` folder, so Python 3 and the unchanged folder layout are required. Installing by cloning into `~/.claude` over an existing setup requires merging histories with `git merge`, and conflicts are possible. The full planning cycle launches several sub-agents and spends more tokens than an ordinary chat.
How to disable. Delete the cloned `.claude` directory in the project or the `skills` and `agents` folders from `~/.claude`. If merged into an existing setup, revert the merge commit.
FAQ
How do I install claude-config for one project and for all?
For one project clone with `git clone https://github.com/solatis/claude-config .claude`, for a fresh global setup into `~/.claude`. If `~/.claude` already has content, add the repository as a remote and run `git merge workflow/main --allow-unrelated-histories`.
What does the working cycle look like?
First exploration and, if needed, `deepthink`. Then a request like "use the planner and write the plan to a file". After `/clear` you ask to execute that plan, and each milestone passes review by the technical writer and the quality reviewer.
Why run /clear after the plan?
All decisions are already written to the plan file, so clearing context loses nothing. The agent starts execution with a clean head and reads only what it needs.
How do deepthink and decision-critic differ?
`deepthink` takes questions where the answer shape is unknown: trade-offs, taxonomies, evaluations. It picks quick or full mode with parallel sub-agents on its own. `decision-critic` stress-tests a decision already made.
What does the refactor skill do?
It analyzes naming, extraction, types, errors, modules, architecture and abstractions in parallel, validates findings against the code and returns a prioritized list. The skill generates no code.
Discussion
No questions or reviews yet.
Related
SkillsA set of slash commands, skills and subagents for Claude Code: review, tests, security, releases, docs and GitHub to Linear sync
A Claude Code environment installer: CLAUDE.md, rules, skills, agents, hooks and two-layer project memory in one script
A set of Claude Code utilities: background skills, expert agents, slash commands and prompt templates for development
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
A ready-made Claude Code setup: agents, slash commands, skills, safety hooks, rules and MCP connections in one install
A Claude Code plugin that generates and maintains CLAUDE.md: a 0 to 100 quality score, a 150-line cap and stale reference checks
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| claude-config (this card) | no | 914 | |
| Superpowers | no | 313,096 | another set of skills and processes for development, with no CLAUDE.md index scheme or token budgets |
| BMad Method | no | 55,033 | a role-based process for team development, while claude-config is built around one planner and per-milestone review |
README in short
The README sets out the author view on why LLM code decays fast and four principles: context hygiene, planning, review cycles and cost-effective delegation. Then come the quick start, a worked example of the cycle and descriptions of individual skills.
README badge
Original SKILL.md
--- name: planner description: Interactive planning and execution for complex tasks. IMMEDIATELY invoke when user asks to use planner. --- ## Activation When this skill activates, IMMEDIATELY invoke the corresponding script. The script IS the workflow. | Mode | Intent | Command | | --------- | ---------------------------------- | ---------------------------------------------------------------------------------------------------------------- | | planning | "plan", "design", "architect" | `<invoke working-dir=".claude/skills/scripts" cmd="python3 -m skills.planner.orchestrator.planner --step 1" />` | | execution | "execute", "implement", "run plan" | `<invoke working-dir=".claude/skills/scripts" cmd="python3 -m skills.planner.orchestrator.executor --step 1" />` |