skill-developer

The Skill Developer skill acts as an authoritative guide for architects and developers building modular, auto-activating workflows within Claude Code. It encapsulates the core architecture for skill development, ensuring that agents adhere to Anthropic's performance standards, including the 500-line rule for documentation and effective progressive disclosure patterns. This toolset is essential for teams aiming to standardize agent behavior, enforce guardrails, and automate knowledge retrieval through custom triggers.

• Full implementation guidance for the two-hook architecture, covering UserPromptSubmit for proactive context injection and stop hooks for error-handling reminders.

• Comprehensive schema management for skill-rules.json, enabling precise control over enforcement levels (block, suggest, warn) and multi-pattern trigger conditions.

• Best practices for file structure, using individual SKILL.md files to house domain-specific logic, constraints, and reference material.

• Support for advanced trigger configurations, including keyword matching, regex-based intent patterns, file path globbing, and content-based validation.

• Debugging utilities to simulate and test trigger activation without full environment deployment.

• Lifecycle management for guardrail skills that block unsafe tool execution until verification steps are met.

• Designed for developers creating domain-specific agents, backend/frontend guidelines, or specialized security guardrails.

• Inputs include the definition of trigger criteria in JSON, markdown-based documentation, and shell-based hook scripts for execution context.

• Users should adhere to the 500-line limit per skill to maintain performance, relying on linked reference files for deeper documentation.

• Always prioritize session tracking to avoid redundant warnings or blocking prompts within a single user conversation.

• When implementing, ensure strict adherence to the defined hook sequence: UserPromptSubmit, PreToolUse, and PostToolUse patterns to maintain agent stability and session awareness.