* fix(shellcheck): declare and assign separately in router.sh (SC2155) * feat(plugin): enhance context management and skill invocation - Update .context-manifest.json to include profile and detailed source information. - Revise context-scout.md to clarify context root discovery and navigation. - Modify commands to replace deprecated skills with updated ones (brainstorming → approach, systematic-debugging → debugger). - Streamline install-context.md for clarity and remove interactive profile selection. - Improve oac-status.md to provide a more comprehensive status report. - Add OAC system paths to session-start.sh for better context management. - Remove obsolete skills and improve documentation for existing commands.
1.9 KiB
1.9 KiB
Skills vs Other Options
Purpose: When to use Skills versus other Claude Code customizations
Comparison Table
| Option | When It Runs | Best For |
|---|---|---|
| Skills | Claude chooses when relevant | Specialized knowledge (review standards, patterns) |
| Slash commands | You type /command |
Reusable prompts (/deploy staging) |
| CLAUDE.md | Every conversation | Project-wide instructions |
| Subagents | Claude delegates or you invoke | Isolated context, different tools |
| Hooks | Specific tool events | Auto-format, logging, protection |
| MCP servers | Claude calls as needed | External tools and data sources |
Key Distinctions
Skills vs Slash Commands
- Skills: Claude decides when to use based on task
- Commands: You explicitly type
/command
Skills vs Subagents
- Skills: Add knowledge to current conversation
- Subagents: Separate context with own tools/permissions
Skills vs MCP
- Skills: Tell Claude how to use tools
- MCP: Provide the tools themselves
Skills vs CLAUDE.md
- Skills: Loaded only when relevant
- CLAUDE.md: Always loaded into every conversation
Decision Guide
Use Skills when:
- Teaching domain-specific knowledge
- Defining review/coding standards
- Want Claude to decide when to apply
Use Subagents when:
- Need tool isolation
- Task produces verbose output
- Want separate context window
Use CLAUDE.md when:
- Rules apply to ALL interactions
- Project-wide coding standards
Related
../concepts/agent-skills.md- Skills overview../concepts/subagents-system.md- Subagents overview../../standards/subagent-structure.md- Subagent patterns
Reference: https://docs.anthropic.com/en/docs/claude-code/skills