TÂCHES Claude Code Resources plugin
The author's personal Claude Code toolbox: meta-prompt and todo commands, skills to create skills, hooks and MCP servers, auditors and a Ralph loop
- 1,986 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The `setup-ralph` skill sets up an autonomous loop where the agent picks tasks, runs checks and makes commits itself
- The `create-hooks` skill creates hooks with shell commands that fire on session events
- Commands write files into the project (`TO-DOS.md`, `./prompts`, plans) and run git
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/taches-claude-code-resourcesDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The `setup-ralph` skill sets up an autonomous loop where the agent picks tasks, runs checks and makes commits itself
- The `create-hooks` skill creates hooks with shell commands that fire on session events
- Commands write files into the project (`TO-DOS.md`, `./prompts`, plans) and run git
This is third-party code. Review the repository files before installing.
What it does
TÂCHES Claude Code Resources is a plugin of commands, skills and subagents that the author assembled for his own work in Claude Code. Commands are grouped by task. Meta-prompting separates analysis from execution: `/create-prompt` builds a prompt with XML structure and saves it to `./prompts`, `/run-prompt` runs one or several prompts in a fresh subtask context, sequentially or in parallel. Notes for later: `/add-to-todos` records a problem with file paths and line numbers in `TO-DOS.md`, `/check-todos` returns to it with full context, `/whats-next` prepares a handoff document to continue in a new session. Another block holds twelve `/consider:*` lens commands (first principles, inversion, the Pareto rule, 5 whys, SWOT, the Eisenhower matrix and others) and eight `/research:*` commands for comparing options, feasibility, competitor overviews and finding open source projects. `/debug` switches on a methodical debugging protocol with hypothesis testing. The skills exist to build your own tooling: `create-agent-skills`, `create-slash-commands`, `create-subagents`, `create-hooks`, `create-mcp-servers` (Python and TypeScript, API research, OAuth) and `create-meta-prompts`. The `create-plans` skill leads a project through a chain from a short brief via a roadmap to `PLAN.md`, `setup-ralph` sets up Geoffrey Huntley's autonomous loop, and three subagents (`skill-auditor`, `slash-command-auditor`, `subagent-auditor`) review created skills, commands and agents.
Who it is for. For Claude Code developers who build their own skills, commands, hooks and MCP servers and want ready working methods for planning, notes and debugging.
Good fit when
- You need to create a new skill, command, hook, subagent or MCP server and have an auditor subagent review the result
- A task fits a split into analysis and execution: a prompt first, then a run in a fresh context
- You spot a problem mid-work and want to record it with paths and lines without losing focus
- A session is overloaded with context and you need a handoff document to continue in a new one
- You have to choose between options or test an idea with a ready thinking model: Pareto, inversion, SWOT, the Eisenhower matrix
Not a fit when
- You need skills for another agent: the set is written for Claude Code, and for others there is only a third-party port for OpenCode
- You need a small targeted tool: the plugin brings dozens of commands and several large skills with long references
- You cannot run unattended autonomous loops: the `setup-ralph` skill sets up exactly that kind of process
- You need only a few individual skills: copying `skills/*` into a project installs the whole folder, and you have to remove the extras yourself
Example request
Run /create-prompt describing the task of adding authentication to the project, then run the resulting prompt with /run-prompt in a separate contextLimitations
The author calls the set his favorite and growing collection of resources, so the contents of the set keep changing: the description speaks of 27 commands and 9 skills, while the repository already has more.
Some commands write files into the project working folder (`TO-DOS.md`, `./prompts`, plan documents), `create-hooks` creates hooks with shell commands, and `setup-ralph` sets up a loop where the agent picks and runs tasks itself, runs checks and makes commits. The plugin manifest version is 1.0.0, and the latest update came in April 2026. MIT license.
How to disable. Remove the plugin through the `/plugin` menu in Claude Code or delete the copied commands from `~/.claude/commands/` and skills from `~/.claude/skills/`. Delete `TO-DOS.md`, `prompts` folders and plan documents in projects by hand. Remove hooks created by the `create-hooks` skill from Claude Code settings.
FAQ
How does the /create-prompt and /run-prompt pair work?
The first command asks clarifying questions if the task description is vague, writes a prompt with XML tags and numbers the file in `./prompts`. The second runs one or several prompts as separate subtasks with a fresh context: by number or part of the name, and with no argument it takes the latest. With several numbers it runs sequentially by default, and `--parallel` turns on parallel execution.
What does /add-to-todos record?
A block in `TO-DOS.md` with a title, timestamp, problem description, file paths with line numbers and a solution hint. `/check-todos` then shows the list and brings back the context of the chosen entry so you can continue from there.
How do skills differ from commands?
Commands expand into a ready prompt when called, like `/debug` or `/consider:pareto`. Skills run an autonomous process with questions and references: they create other skills, hooks, MCP servers and plans. The `/create-*` commands are wrappers that launch the skills of the same names.
What do the auditors check?
The subagents `skill-auditor`, `slash-command-auditor` and `subagent-auditor` read the kit's references and review created skills, commands and agents: XML markup, YAML description, progressive disclosure, error handling and example quality. They only read files and report findings with line numbers. They are called with `/audit-skill`, `/audit-slash-command` and `/audit-subagent`, and `/heal-skill` fixes a skill after a failed run.
Can it be installed without the plugin?
Yes. After `git clone`, copy `commands/*` into `~/.claude/commands/` and `skills/*` into `~/.claude/skills/`. Commands install globally, while project data such as prompts and notes stays in each project's working folder.
Discussion
No questions or reviews yet.
Related
PluginsA large kit for Claude Code: subagents, skills, commands, hooks, rules, templates and MCP configs in one plugin marketplace
A learning kit for Claude Code: feature walkthroughs plus copy-paste templates for commands, skills, memory, hooks and subagents
A Claude Code plugin marketplace with ready sets of agents, slash commands, hooks and skills, plus a search site
A set of skills, slash commands and agents for Claude Code that helps you build your own skills, agents, hooks, slash commands and prompts
Workflows for Claude Code: /do, /omo, /bmad-pilot and more commands, with Codex, Gemini or OpenCode writing the code through codeagent-wrapper
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| TÂCHES Claude Code Resources (this entry) | no | 1,986 | |
| SuperClaude Framework | no | 23,911 | /sc slash commands, agents and behavior modes for all stages of development, without extension-building skills |
| Claude Code Skills & Agents Factory | no | 893 | a workshop for building your own skills, agents and hooks through questions from the factory-guide agent |
| Ralph | no | 22,053 | just a bash loop that restarts Claude Code or Amp with a clean context over the tasks from a PRD |
README in short
The description starts with the author's short philosophy and a list of three groups: commands, skills and agents. It continues with two ways to install, commands by theme (meta-prompting, notes for later, context handoff, creating and auditing extensions, self-repair, thinking models, deep debugging), agents and a description of each skill. It ends with a recommended workflow: how to run a project, how to accumulate domain knowledge and how to build your own pipelines where one Claude hands work to another. The repository holds three documents: on context handoff, meta-prompting and notes for later.