# Create Production AI Agents

Teams need consistent agents that encode domain rules and workflows. This skill guides Claude, Codex, and Claude Code through agent design, SDK setup, testing, and packaging.

## Install

```bash
npx skillstore add dnyoussef/agent-creator
```

## Metadata

- Status: approved
- Slug: dnyoussef-agent-creator
- Skillstore revision: r1
- Version status: missing
- Tree hash: 9e90920f08b7e8b390ce0e8f8fa01e1930979229a86cf48f50c30ed025c0d8d1
- Author: DNYoussef
- GitHub username: DNYoussef
- License: MIT
- Repository: https://github.com/DNYoussef/ai-chrome-extension/tree/main/.claude/skills/agent-creator
- Ref: c579efd02184f1128d6741d16e6c8d951d7d4208
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 68
- Public page: https://skillstore.pages.dev/skills/dnyoussef-agent-creator
- Manifest: https://skillstore.pages.dev/api/skills/dnyoussef-agent-creator/manifest

## Capabilities

- Provides a four-phase SOP for agent research, specification, architecture, and technical enhancement.
- Defines system prompt sections for identity, commands, tools, workflows, guardrails, and success criteria.
- Includes TypeScript and Python Claude Agent SDK implementation examples.
- Documents MCP integration, memory namespace patterns, and cross-agent coordination concepts.
- Guides validation with typical cases, edge cases, error cases, integration cases, and performance checks.

## Use Cases

- Build a specialist coding agent: Create an agent that understands a project stack, coding standards, workflows, and validation steps.
- Standardize operational agents: Turn repeated business workflows into consistent agents with documented inputs, outputs, and guardrails.
- Design coordinated agent systems: Plan agents that share memory, coordinate tasks, and report progress through defined MCP patterns.

## Prompt Templates

### Create a basic agent

```
Use agent-creator to design a specialist agent for [role]. Include purpose, boundaries, core tasks, required knowledge, and success criteria.
```

### Improve an existing agent

```
Review this agent prompt with agent-creator. Strengthen its domain knowledge, failure modes, validation checks, and output format.
```

### Prepare an SDK-ready agent

```
Use agent-creator to convert this specification into an SDK-ready agent plan with tools, permissions, tests, and packaging notes.
```

### Design a multi-agent workflow

```
Use agent-creator to define a multi-agent workflow for [process]. Include agent roles, handoffs, memory keys, escalation rules, and validation gates.
```

## Limitations

- Does not implement a complete runnable agent by itself.
- Requires users to verify current SDK APIs, package versions, and local environment settings.
- Needs security review before enabling Bash, Write, MCP servers, or broad permission modes.
- Works best when domain research and example tasks are already available.

## Best Practices

- Start with domain research before writing the system prompt.
- Define guardrails, failure modes, and validation checks before implementation.
- Use minimal tool permissions and review any command-capable integration.

## Anti Patterns

- Do not create generic agents without domain-specific examples and boundaries.
- Do not enable broad Bash, Write, or bypass permissions without explicit review.
- Do not skip testing for edge cases, error handling, and output consistency.

## Security Audit

- Audited at: 2026-07-05T12:10:07.042\+00:00
- Summary: All static findings are false positives caused by Markdown backticks, code fences, or DOT graph edges rather than executable behavior. The remaining risk is contextual: SDK templates encourage broad Write, Bash, npx, and permission-bypass patterns without enough safety guidance.

## Stats

- Views: 262
- Downloads: 19
- Favorites: 0
- Popularity score: 0