| name | using-agent-harness |
| description | Use when starting any conversation โ establishes skill discovery and requires Skill tool invocation before ANY response. |
| disable-model-invocation | true |
If you were dispatched as a subagent to execute a specific task, skip this skill.
The Rule
Invoke relevant or requested skills BEFORE any response or action. Even a 1% chance a skill applies means invoke it to check.
Instruction Priority
- User's explicit instructions (CLAUDE.md, AGENTS.md, direct requests) โ highest priority
- Agent Harness skills โ override default system behavior where they conflict
- Default system prompt โ lowest priority
How to Access Skills
- Claude Code: Use the
Skill tool. Never Read skill files directly.
- Copilot CLI: Use the
skill tool (auto-discovered from plugins).
- Gemini CLI: Use the
activate_skill tool.
Codex: subagent setup, environment detection, and App finishing notes in references/codex-tools.md.
Skill Selection
When multiple skills could apply, use this order:
- Process skills first (brainstorming, debugging) โ they determine HOW to approach the task.
- Implementation skills second (design, mcp-builder) โ they guide execution.
"Let's build X" โ brainstorming first, then implementation skills.
"Fix this bug" โ debugging first, then domain-specific skills.
Skill Types
- Rigid (TDD, debugging): Follow exactly. Don't adapt away discipline.
- Flexible (patterns): Adapt principles to context.
The skill itself tells you which.
Anti-Rationalization
These thoughts mean STOP โ you're rationalizing:
- "This is just a simple question" โ Questions are tasks. Check for skills.
- "I need more context first" โ Skill check comes BEFORE clarifying questions.
- "Let me explore the codebase first" โ Skills tell you HOW to explore. Check first.
- "This doesn't need a formal skill" โ If a skill exists, use it.
- "I remember this skill" โ Skills evolve. Read current version.
User instructions say WHAT, not HOW. "Add X" or "Fix Y" doesn't mean skip workflows.
