refactor: organize commands, skills, and agents with clear usage criteria

[p-iknow]

Context

We should find proper way to use command, skills, agents and update to that way.

Current state analysis:

• Commands (6): branch.md, commit.md, pr.md, issue-task.md, plan.md, make-command.md
• Skills (2): meta-skill-creator/, meta-agent-creator/
• Agents (0): Only creator skills exist, no actual agents defined

Issues identified:

1. Inconsistent complexity in commands

   • commit.md: 65 lines (appropriate for command)
   • make-command.md: 654 lines (too complex for command, should be skill?)
   • issue-task.md: 388 lines (borderline)

2. No clear decision criteria for when to use:

   • Commands (simple, triggered by /slash)
   • Skills (comprehensive guides with scripts, references, assets)
   • Agents (specialized subagents for delegation)

3. Documentation gap - .claude/AGENTS.md exists but doesn't provide clear guidance on choosing between commands vs skills vs agents

Requirement

Organized skills, commands, and agents with clear usage criteria and proper structure.

Solution

1. Define clear criteria for each type:

   • Commands: <200 lines, simple workflows, triggered by /slash
   • Skills: >200 lines OR needs scripts/references/assets, comprehensive guides
   • Agents: Specialized subagents for delegation in multi-agent orchestration

2. Refactor oversized commands:

   • Convert make-command.md (654 lines) into a skill
   • Evaluate issue-task.md (388 lines) for potential skill conversion

3. Update documentation:

   • Add decision tree to .claude/AGENTS.md
   • Document when to use commands vs skills vs agents

Verification

Reference

• .claude/commands/*.md (all 6 commands)
• .claude/skills/*/SKILL.md (both skills)
• .claude/AGENTS.md (documentation)