How this agent operates — its isolation, permissions, and tool access model
Agent reference
oh-my-claudecode:agents/agentsThe summary Claude sees when deciding whether to delegate to this agent
<!-- Parent: ../AGENTS.md --> <!-- Generated: 2026-01-28 | Updated: 2026-01-28 --> Markdown prompt templates for all 32 agents in oh-my-claudecode. This directory contains the prompt templates that define agent behavior. Each file is a markdown document with YAML frontmatter for metadata, loaded dynamically by `src/agents/definitions.ts`. | File | Agent | Model | Purpose | |------|-------|-----...
Markdown prompt templates for all 32 agents in oh-my-claudecode.
This directory contains the prompt templates that define agent behavior. Each file is a markdown document with YAML frontmatter for metadata, loaded dynamically by src/agents/definitions.ts.
| File | Agent | Model | Purpose |
|---|---|---|---|
architect.md | architect | opus | Architecture, debugging, root cause analysis |
architect-medium.md | architect-medium | sonnet | Moderate analysis tasks |
architect-low.md | architect-low | haiku | Quick code questions |
executor.md | executor | sonnet | Focused task implementation |
executor-high.md | executor-high | opus | Complex multi-file changes |
executor-low.md | executor-low | haiku | Simple single-file tasks |
explore.md | explore | haiku | Fast codebase search |
explore-medium.md | explore-medium | sonnet | Thorough search with reasoning |
explore-high.md | explore-high | opus | Architectural discovery |
designer.md | designer | sonnet | UI/UX, component design |
designer-high.md | designer-high | opus | Complex UI architecture |
designer-low.md | designer-low | haiku | Simple styling tweaks |
researcher.md | researcher | sonnet | Documentation research |
researcher-low.md | researcher-low | haiku | Quick doc lookups |
writer.md | writer | haiku | Technical documentation |
vision.md | vision | sonnet | Image/screenshot analysis |
critic.md | critic | opus | Critical plan review |
analyst.md | analyst | opus | Pre-planning requirements |
planner.md | planner | opus | Strategic planning |
qa-tester.md | qa-tester | sonnet | Interactive CLI testing |
qa-tester-high.md | qa-tester-high | opus | Production-ready QA |
scientist.md | scientist | sonnet | Data analysis |
scientist-high.md | scientist-high | opus | Complex ML/research |
scientist-low.md | scientist-low | haiku | Quick data inspection |
security-reviewer.md | security-reviewer | opus | Security audits |
security-reviewer-low.md | security-reviewer-low | haiku | Quick security scans |
build-fixer.md | build-fixer | sonnet | Build error resolution |
build-fixer-low.md | build-fixer-low | haiku | Simple type errors |
tdd-guide.md | tdd-guide | sonnet | TDD workflow |
tdd-guide-low.md | tdd-guide-low | haiku | Test suggestions |
code-reviewer.md | code-reviewer | opus | Expert code review |
code-reviewer-low.md | code-reviewer-low | haiku | Quick code checks |
Each file follows this structure:
---
name: agent-name
description: Brief description of what this agent does
model: opus | sonnet | haiku
tools: [Read, Glob, Grep, ...]
---
# Agent Name
## Role
What this agent is and its expertise.
## Instructions
Detailed instructions for how the agent should behave.
## Constraints
What the agent should NOT do.
## Output Format
How results should be formatted.
new-agent.md with YAML frontmattersrc/agents/definitions.tsFor model routing, create variants with complexity-appropriate instructions:
| Tier | File Suffix | Instructions Focus |
|---|---|---|
| LOW (Haiku) | -low.md | Quick, simple tasks, minimal reasoning |
| MEDIUM (Sonnet) | Base file or -medium.md | Standard complexity |
| HIGH (Opus) | -high.md | Complex reasoning, deep analysis |
Frontmatter parsing (in definitions.ts):
function loadAgentPrompt(agentName: string): string {
const content = readFileSync(`agents/${agentName}.md`, 'utf-8');
const match = content.match(/^---[\s\S]*?---\s*([\s\S]*)$/);
return match ? match[1].trim() : content.trim();
}
Tool assignment by agent type:
[Read, Glob, Grep][Read, Glob, Grep, WebSearch, WebFetch][Read, Glob, Grep, Edit, Write, Bash, TodoWrite][Read, Glob, Grep, Bash, python_repl]Agent prompts are tested via integration tests that spawn agents and verify behavior.
src/agents/definitions.tsskills/None - pure markdown files.
| Category | Agents | Common Tools |
|---|---|---|
| Analysis | architect, architect-medium, architect-low | Read, Glob, Grep, WebSearch |
| Execution | executor, executor-low, executor-high | Edit, Write, Bash, TodoWrite |
| Search | explore, explore-medium, explore-high | Read, Glob, Grep |
| Research | researcher, researcher-low | WebSearch, WebFetch |
| Frontend | designer, designer-low, designer-high | Edit, Write, Bash |
| Docs | writer | Edit, Write |
| Visual | vision | Read, Glob, Grep |
| Planning | planner, analyst, critic | Read, Glob, Grep |
| Testing | qa-tester, qa-tester-high | Bash, TodoWrite |
| Security | security-reviewer, security-reviewer-low | Read, Grep, Bash |
| Build | build-fixer, build-fixer-low | Edit, Write, Bash |
| TDD | tdd-guide, tdd-guide-low | Edit, Write, Bash |
| Review | code-reviewer, code-reviewer-low | Read, Grep, Bash |
| Data | scientist, scientist-low, scientist-high | Bash, python_repl |
npx claudepluginhub letrandat/oh-my-claudecodeSurgical single-file editor that fixes typos, rewrites functions, renames, and removes comments with minimal diff. Refuses changes spanning 3+ files or new abstractions. Returns a compact diff receipt.
RuView model specialist who trains camera-free WiFlow pose, camera-supervised pose, RuVector embeddings, and domain-generalized SNN adapters, then publishes to Hugging Face. Delegate all model-building tasks here.