agent-design · diff
git:20260705.ba8ee52 to git:20260729.3b91eed
7 added, 1 removed. Audit A to A.
---
name: agent-design
- description: "Design AI agents with recommended patterns and architectures. Use when: choosing between workflow vs agent patterns, structuring orchestrator/subagent pipelines, or writing a Claude Code agent template."
+ description: "Use when choosing between workflow vs agent patterns, structuring orchestrator/subagent pipelines, or writing a Claude Code agent template."
allowed-tools: Read, Write
---
+
+ <objective>
+ Agent Design covers the architectural decisions behind building a high-performance AI agent: the workflow-vs-agent distinction (code-orchestrated vs LLM-directed control), the minimal agent structure (identity, capabilities, tools, constraints, workflow), and the "fresh eyes" principle -- each sub-agent gets only the context it needs, never the full orchestrator history.
+
+ It routes to four references depending on the task: `patterns.md` for architecture choice (Single Agent, Orchestrator, Pipeline, Network, Supervisor, Hierarchical, Meta-Prompting), `workflows.md` for implementing APEX/TDD/Explore-Plan-Code/Review/Debug patterns inside an agent, `templates.md` for the full production frontmatter and Claude Code agent template, and `anti-patterns.md` for reviewing a design against common mistakes.
+ </objective>
# Agent Design
Skill for designing high-performance AI agents following 2025 patterns.
## References
- [patterns.md](references/patterns.md) - Load when: choosing an agent architecture (Single Agent, Agent+Tools, Orchestrator, Pipeline, Network, Supervisor, Hierarchical, Meta-Prompting) or comparing their trade-offs
- [workflows.md](references/workflows.md) - Load when: implementing APEX, TDD, Explore-Plan-Code, Code Review, or Debugging workflows for an agent
- [templates.md](references/templates.md) - Load when: writing a full agent definition (production frontmatter + sections) or a Claude Code agent template
- [anti-patterns.md](references/anti-patterns.md) - Load when: reviewing an agent design for common mistakes (omniscient agent, implicit instructions, missing error handling)
## Fundamental Distinction
### Workflows vs Agents
| Type | Control | When to use |
|------|---------|-------------|
| **Workflow** | Code orchestrates LLM | Predictable tasks, need for control |
| **Agent** | LLM directs its actions | Flexibility, adaptive decisions |
**Golden rule:** Start simple, add complexity if necessary.
## Minimal Agent Structure
```yaml
Agent:
identity: Who am I?
capabilities: What can I do?
tools: What tools do I have?
constraints: What are my limits?
workflow: How should I proceed?
```
For the complete production structure and the Claude Code agent template, see [templates.md](references/templates.md).
## Fresh Eyes Principle
**Key 2025 concept:** Each sub-agent must have a "fresh" context.
```
❌ Bad: Pass entire history to each sub-agent
✅ Good: Give only necessary information
Orchestrator:
- Keeps complete history
- Extracts relevant context for each sub-agent
- Synthesizes results
```
## Design Checklist
### Before creating an agent
- [ ] Is the objective clear?
- [ ] Would a simple workflow suffice?
- [ ] What tools are needed?
- [ ] What guardrails are required?
### During design
- [ ] Is identity well defined?
- [ ] Is workflow explicit?
- [ ] Are error cases handled?
- [ ] Are examples relevant?
### After creation
- [ ] Standard case tests?
- [ ] Edge case tests?
- [ ] Security tests (jailbreak)?
- [ ] Acceptable performance?
For anti-patterns to avoid during design, see [anti-patterns.md](references/anti-patterns.md).
## Forbidden
- Never create an agent without explicit workflow
- Never give access to all tools without necessity
- Never ignore the principle of least privilege
- Never forget security guardrails