claude-code-agents-and-skills - SKILL.md Agent Skill

name: claude-code-agents-and-skills description: Deploy specialized Claude Code agents with isolated context windows and reusable skills across instances. Trigger when building complex projects requiring specialized capabilities like marketing copy, UX design, or multi-step planning workflows that need dedicated context isolation or cross-project skill reuse. source: 10 Claude Code tips I wish I knew from the start + https://youtu.be/BEIulrjHzMI?si=JiZ_XdfyLiw9nCGP generated_by: SkillJack (skilljacked.com)

Claude Code Agents and Skills Deployment

Deploy specialized agents with isolated context windows and create reusable skills for complex project workflows.

Quick Start

Create agent: New agent → Choose tools/model → Save. Invoke: "Use the agent [name] to [task]" Create skill: Type "create a skill that [purpose]" or /skills to manage existing ones.

Core Workflow

Agent Creation and Deployment

WHEN building complex projects requiring specialized capabilities:

Create new agent with specific tools (read-only/full access)
Select model power level (Haiku/Sonnet/Opus)
Assign color for visual identification
Save agent for reuse

Agent Invocation

WHEN you need specialized expertise in existing chat:

Type "use the agent [agent-name]" followed by task
Agent works in isolated context window (different color display)
Agent returns results to main conversation
Multiple agents can work simultaneously

Skills Creation and Management

WHEN building reusable capabilities across projects:

Use /skills to list available skills
Create new: "create a skill that [specific purpose using best practices]"
Claude creates markdown file in skills directory
Skills work within current context window (unlike agents)
Skills can be stacked together in single workflow

Techniques

Agent Specialization Patterns

Marketing copy agent for landing pages/headlines
UX/UI agent for design decisions
Planning agent for project architecture
Ask Claude: "what kind of agents do you recommend I create?"

Skill Stacking

Combine multiple skills: "Use title generator skill and script generator skill together"
Skills invoke separately but work in unified workflow
Each skill maintains its specialized knowledge base

Ultra Think Mode Enhancement

Type "ultra think" before any prompt
Changes to rainbow color display
Uses more tokens for deeper reasoning
Works with both agents and skills

Anti-Patterns

NEVER create agents without clear specialization purpose NEVER confuse agents (isolated context) with skills (current context)
NEVER use ultra think mode unnecessarily (token intensive) NEVER forget agents can work simultaneously - leverage parallelization

Edge Cases & Error Handling

Agent Context Isolation: Agents don't retain conversation history from main chat Skill Persistence: Skills persist across all Claude Code instances once created Model Selection: Choose agent model based on task complexity (Opus for complex reasoning) Tool Access: Match agent permissions to required capabilities

Bundled Resources Plan

scripts/agent-templates.md - Pre-configured agent setups for common use cases
skills/skill-library.md - Curated collection of proven skill patterns
references/context-management.md - Best practices for agent vs skill selection