0chan-smc-skill-developer_skill

Overview

This skill is a meta-tool for creating and managing Claude Code skills, hooks, agents, and commands. It helps you define activation triggers, configure enforcement and priorities, and maintain the infrastructure that runs skill logic. Use it to standardize how skills are organized, discovered, and executed across a codebase.

How this skill works

The skill inspects trigger configuration (keyword and intent patterns) and file-based triggers (path globs, exclusions, and content regex) to decide when a skill should activate. It generates and validates skill entries, sets enforcement modes (suggest, warn, block), and assigns priority levels so the runtime can suggest or enforce actions. It also helps manage hook and agent wiring for command execution and lifecycle events.

When to use it

• Creating or scaffolding a new skill and its resources
• Configuring skill activation rules for prompt or file-based triggers
• Setting enforcement behavior to suggest, warn, or block
• Adding or tuning hooks and agent command definitions
• Testing and debugging why a skill did or did not activate

Best practices

• Organize each skill with a concise main guideline file and optional resource files for examples
• Use specific path patterns and sensible exclusions to avoid over-triggering
• Prefer intent regex for flexible prompt matching and keywords for simple cases
• Set enforcement appropriately: use block for guardrails, suggest for guidance, warn for soft constraints
• Document triggers, priority, and customization notes alongside the skill files

Example use cases

• Add a frontend guidance skill that triggers on app/**/*.tsx files and Next.js import patterns
• Create a guardrail that blocks commits if security-related content patterns are detected
• Define an error-tracking skill that fires when Sentry usage or captureException appears in service files
• Add slash commands and hooks to run automation tasks via agents when a skill activates
• Tune intentPatterns to respond to developer prompts like “how to implement component”