plugins/plugin-master/skills/plugin-master/SKILL.md
Complete guide to Claude Code plugin development, architecture, directory layout, components, and marketplace publishing. PROACTIVELY activate for: (1) creating a plugin from scratch, (2) building or scaffolding a plugin, (3) writing plugin.json, (4) adding commands/agents/skills/hooks/MCP servers, (5) packaging code as a plugin, (6) publishing to a marketplace, (7) validating plugin structure, (8) marketplace.json registration, (9) cross-platform compatibility, (10) version and metadata sync. Provides: plugin schema, layouts, workflows, and publishing steps.
npx skillsauth add JosiahSiegel/claude-plugin-marketplace plugin-masterInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
| Component | Location | Required |
|-----------|----------|----------|
| Plugin manifest | .claude-plugin/plugin.json | Yes |
| Commands | commands/*.md | No (auto-discovered) |
| Agents | agents/*.md | No (auto-discovered) |
| Skills | skills/*/SKILL.md | No (auto-discovered) |
| Hooks | hooks/hooks.json | No |
| MCP Servers | .mcp.json | No |
| Task | Action |
|------|--------|
| Create plugin | Ask: "Create a plugin for X" |
| Validate plugin | Run: /validate-plugin |
| Install from marketplace | /plugin marketplace add user/repo then /plugin install name@user |
plugin-name/
├── .claude-plugin/
│ └── plugin.json # MUST be inside .claude-plugin/
├── agents/
│ └── domain-expert.md
├── commands/
├── skills/
│ └── skill-name/
│ ├── SKILL.md
│ ├── references/
│ └── examples/
└── README.md
{
"name": "plugin-name",
"version": "1.0.0",
"description": "Complete [domain] expertise. PROACTIVELY activate for: (1) ...",
"author": {
"name": "Author Name",
"email": "[email protected]"
},
"license": "MIT",
"keywords": ["keyword1", "keyword2"]
}
Validation Rules:
author MUST be an object { "name": "..." } - NOT a stringversion MUST be a string "1.0.0" - NOT a numberkeywords MUST be an array ["word1", "word2"] - NOT a stringagents, skills, slashCommands - these are auto-discoveredALL markdown files in agents/, commands/, skills/ MUST begin with frontmatter:
---
description: Brief description of what this component does
---
# Content...
Without frontmatter, components will NOT load.
{domain}-expertNaming Standard:
docker-master → agent named docker-expertterraform-master → agent named terraform-expertSkills use three-tier loading:
This enables unbounded capacity without context bloat.
Before creating files, check:
# Check if in marketplace repo
if [[ -f .claude-plugin/marketplace.json ]]; then
PLUGIN_DIR="plugins/PLUGIN_NAME"
else
PLUGIN_DIR="PLUGIN_NAME"
fi
# Get author from git config
AUTHOR_NAME=$(git config user.name)
AUTHOR_EMAIL=$(git config user.email)
mkdir -p $PLUGIN_DIR/.claude-plugin
mkdir -p $PLUGIN_DIR/agents
mkdir -p $PLUGIN_DIR/skills/domain-knowledge
If the plugin ships any vendored, derived, or licensed third-party content, create NOTICES.md at the plugin root before registering in the marketplace. Treat it as a first-class shipping artifact alongside plugin.json and README.md, not as doc polish. See references/publishing-guide.md ("Licensed / Vendored / Derived Content" checklist) for the structural integrity, license-text-preservation, and cross-reference requirements.
If the plugin contains no third-party content, skip this step.
CRITICAL: If .claude-plugin/marketplace.json exists at repo root, you MUST add the plugin:
{
"plugins": [
{
"name": "plugin-name",
"source": "./plugins/plugin-name",
"description": "Same as plugin.json description",
"version": "1.0.0",
"author": { "name": "Author" },
"keywords": ["same", "as", "plugin.json"]
}
]
}
User-initiated slash commands in commands/*.md:
---
description: What this command does
---
# Command Name
Instructions for Claude to execute...
Autonomous subagents in agents/*.md:
---
name: agent-name
description: |
Brief role summary. PROACTIVELY activate for: (1) trigger, (2) trigger, ..., (N) trigger. Provides: capability list.
# Optional. Include 3-5 <example> blocks ONLY when the agent body
# exceeds 2,500 words. Lean orchestrators omit them by design.
# See agent-development "Example-block requirement by agent body size".
model: inherit
color: blue
---
System prompt for agent...
Dynamic knowledge in skills/skill-name/SKILL.md:
---
name: skill-name
description: When to use this skill...
---
# Skill content with progressive disclosure...
Event automation in hooks/hooks.json:
{
"PostToolUse": [{
"matcher": "Write|Edit",
"hooks": [{
"type": "command",
"command": "${CLAUDE_PLUGIN_ROOT}/scripts/lint.sh"
}]
}]
}
Events: PreToolUse, PostToolUse, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification, Stop, SubagentStop
kebab-case (e.g., code-review-helper)review-pr, run-tests)code-reviewer, test-generator)api-design, error-handling)Use ${CLAUDE_PLUGIN_ROOT} for all internal paths:
"command": "${CLAUDE_PLUGIN_ROOT}/scripts/run.sh"
Never use hardcoded absolute paths.
$MSYSTEM, use GitHub method| Issue | Solution |
|-------|----------|
| Plugin not loading | Check plugin.json is in .claude-plugin/ |
| Commands missing | Verify frontmatter has description field |
| Agent not triggering | Check description has PROACTIVELY activate for: enumeration. Add 3-5 <example> blocks only if agent body > 2,500 words — see agent-development SKILL.md "Example-block requirement by agent body size". Lean orchestrators are exempt. |
| Marketplace not found | Ensure repo is public, check path in marketplace.json |
For detailed information, see:
references/manifest-reference.md - Complete plugin.json fieldsreferences/component-patterns.md - Advanced component patternsreferences/publishing-guide.md - Marketplace publishing detailsexamples/minimal-plugin.md - Simplest working pluginexamples/full-plugin.md - Complete plugin with all featuresdevelopment
This skill should be used when the user asks to train, debug, scale, or improve ML models. PROACTIVELY activate for: (1) PyTorch, TensorFlow/Keras, JAX, Flax, Hugging Face Trainer/Accelerate training loops, (2) distributed training, DDP/FSDP/DeepSpeed, TPU/GPU setup, (3) mixed precision AMP/bf16, gradient accumulation, checkpointing, seeding, (4) overfitting, imbalance, loss functions, regularization, LR schedules, warmup, (5) memory optimization, gradient checkpointing, offloading, quantization-aware training. Provides: reproducible training best practices across deep learning and classical ML.
development
This skill should be used when the user asks to productionize, track, version, govern, monitor, or automate ML systems. PROACTIVELY activate for: (1) MLflow, Weights & Biases, Neptune, Comet, ClearML experiment tracking, (2) model registry, model versioning, artifact lineage, reproducibility, (3) Kubeflow, SageMaker Pipelines, Vertex AI Pipelines, Azure ML pipelines, Databricks workflows, (4) CI/CD, continuous training/evaluation, A/B tests, canary/shadow deployments, (5) drift detection, model monitoring, data validation, responsible AI governance. Provides: end-to-end MLOps architecture and operational safeguards.
development
This skill should be used when the user asks to optimize, export, serve, compress, or accelerate ML inference. PROACTIVELY activate for: (1) latency, throughput, p95/p99, batching, concurrency, KV cache, memory, or cost issues, (2) quantization INT8/INT4, GPTQ, AWQ, bitsandbytes, pruning, sparsity, distillation, (3) ONNX export, ONNX Runtime, TensorRT, TorchScript, torch.compile, XLA, OpenVINO, Core ML, TFLite, (4) Triton, TorchServe, TF Serving, BentoML, Seldon, KServe configuration, (5) edge deployment, CPU/GPU/TPU/Inferentia serving. Provides: hardware-aware inference optimization and safe benchmarking.
testing
This skill should be used when the user asks to tune hyperparameters, run sweeps, optimize search spaces, or use AutoML. PROACTIVELY activate for: (1) Optuna, Ray Tune, FLAML, AutoGluon, Hyperopt, Nevergrad, KerasTuner, W&B sweeps, (2) grid search, random search, Bayesian optimization, TPE, Gaussian processes, evolutionary search, (3) ASHA, Hyperband, successive halving, multi-fidelity optimization, population-based training, (4) learning-rate finder, batch-size search, early stopping, pruning, (5) reproducible sweep design and experiment analysis. Provides: budget-aware hyperparameter search strategy.