TÂCHES Principled
Version: 0.4.0
A principle-based Claude Code plugin for building skills, subagents, and project plans.
For when you keep pasting the same instructions into chat, or when CLAUDE.md has grown into a procedure. Each skill teaches you to build better extensions — not by giving you templates and checklists, but by giving you the principles behind them.
Quick Start
# Install from GitHub marketplace
/plugin marketplace add Git-Fg/taches-principled
/plugin install taches-principled
# Or manual install
cp -r skills/* commands/* agents/* ~/.claude/
# Create a new skill
/skill create-skills
# Plan a project phase
/skill create-plans
# Create an executable prompt
/skill create-prompts
# Run a prompt via sub-task
/skill execute-prompts
# Debug an issue
/debug <issue description>
# Create a handoff for the next session
/whats-next
# Simplify code
/simplify [file-pattern]
Note: Skills are invoked via /skill <name> or by description routing — they don't create individual slash commands. The command shortcuts above (/debug, /simplify, /whats-next) are the only flat slash commands. All other skills load on description match.
What's Inside
22 Skills
Skills load on demand and give Claude domain expertise without bloating every conversation.
| Skill | When to Use | Enhancements |
|---|
| create-skills | Building new skills or improving existing ones | Policy/Mechanism, Anti-Patterns, Thresholds |
| create-subagents | Creating specialized agents or configuring the Task tool | Policy/Mechanism, Anti-Patterns, Thresholds |
| create-plans | Planning projects, phases, or features for Claude to build | Policy/Mechanism, Anti-Patterns, Thresholds |
| create-prompts | Creating executable prompts for Claude Code sessions | Policy/Mechanism, Anti-Patterns, Thresholds |
| execute-prompts | Executing prompts via delegated sub-tasks | Policy/Mechanism, Anti-Patterns, Thresholds |
| execute-plans | Executing PLAN.md files via parallel subagent orchestration | Policy/Mechanism, Anti-Patterns, Thresholds |
| subagent-orchestration | Orchestrating parallel subagents for delegated work with self-review loops | RACE Framework, 5 Parallel Patterns, Three Automation Layers |
| code-simplify | Simplifying code post-hoc for clarity and maintainability | Pipeline, Policy/Mechanism, Anti-Patterns, Thresholds |
| add-task | Capturing a task idea for structured development | Standardized folder structure, type classification |
| analyse | Analyzing problems with structured investigation methods | Fishbone, A3, Five Whys |
| analyse-problem | Comprehensive A3 problem analysis with root cause and action plan | A3 one-page format |
| code-review | Reviewing code diffs for correctness bugs | Multi-effort review levels |
| ideation | Generating and refining ideas systematically | Creative sampling, collaborative brainstorming |
| implement-task | Implementing refined task specs with LLM-as-Judge verification | Quality-gated implementation |
| kaizen | Continuous improvement with multiple Kaizen methods | Gemba Walk, Value Stream, Muda |
| plan-do-check-act | Iterative experimentation cycles for systematic improvement | PDCA cycle |
| plan-task | Refining draft specs into implementation-ready tasks | Multi-phase refinement, quality gates |
| reflexion | Reflecting on past work to extract lasting insights | Agentic Context Engineering |
| root-cause-analysis | Systematic root cause investigation using structured methods | Multiple RCA methods |
| root-cause-tracing | Debugging by tracing backward through call stacks | Systematic backtracing with instrumentation |
| update-docs | Maintaining project documentation via multi-agent workflow | Tech-writer agents, quality review |
| write-concisely | Clear, professional writing for human-readable docs | Writing rules and standards |
3 Commands
Slash commands for quick, focused workflows.
| Command | What It Does |
|---|
/debug | Apply systematic debugging methodology |
/whats-next | Create a handoff for a fresh session |
/simplify | Simplify and refine recently modified code |
7 Agents
Specialized agents for quality, review, and evaluation work.
| Agent | Purpose |
|---|
| code-reviewer | Reviews code for issues that matter |
| prompt-engineer | Reviews prompts for clarity and effectiveness |
| skill-auditor | Reviews skills for clarity and routing |
| subagent-auditor | Reviews subagents for effectiveness |
| grader | Evaluates skill teaching effectiveness on 4 dimensions |
| comparator | Compares skill versions for delta analysis |
| analyzer | Synthesizes evaluation results into improvement plans |
6 Separate Plugins
