# Build Reliable AI Agent Tools

Poorly defined tools cause invalid calls, silent failures, and wasted tokens. This skill guides schema design, validation, error handling, MCP integration, and parallel execution.

## Install

```bash
npx skillstore add sickn33/agent-tool-builder
```

## Metadata

- Status: approved
- Slug: sickn33-agent-tool-builder
- Skillstore revision: r2
- Version status: missing
- Tree hash: d96fe40f603aa3113b6483f780a428bcd3aed1fbe56ab55b2942b8fe0674f481
- Author: sickn33
- GitHub username: sickn33
- License: MIT
- Repository: https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/agent-tool-builder
- Ref: ebdfe608f5de2b66ff37ab4af12af8ac4f5e8006
- 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, network
- Quality score: 77
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/sickn33-agent-tool-builder
- Manifest: https://skillstore.pages.dev/api/skills/sickn33-agent-tool-builder/manifest

## Capabilities

- Defines clear tool names, descriptions, parameters, required fields, enums, and strict schema constraints.
- Adds realistic input examples for nested or format-sensitive tool arguments.
- Designs structured errors that help agents identify failures and recover.
- Shows MCP server patterns for listing tools, handling calls, and returning results.
- Explains Anthropic and TypeScript tool runners for managed function-calling loops.
- Organizes independent tool calls for parallel execution and unified result handling.

## Use Cases

- Define a First Agent Tool: Create a clear schema, parameter descriptions, and recoverable errors for a new function-calling integration.
- Standardize an MCP Service: Review tool discovery, call handling, response formatting, and access controls before sharing an MCP server.
- Improve Tool Reliability: Diagnose invalid calls, ambiguous descriptions, silent failures, and inefficient sequential execution in an existing agent.

## Prompt Templates

### Draft a Tool Schema

```
Design a tool for [task]. Define its name, purpose, parameters, required fields, constraints, return text, and likely validation errors.
```

### Improve an Existing Tool

```
Review this tool definition: [definition]. Identify ambiguous descriptions, missing constraints, weak examples, and unrecoverable errors. Provide a revised design.
```

### Plan an MCP Server

```
Plan an MCP server for [system]. Define a small tool set, schemas, permissions, error behavior, transport, observability, and integration tests.
```

### Audit a Production Tool Loop

```
Audit this agent tool loop: [workflow]. Assess selection accuracy, validation, retries, timeouts, parallel calls, result ordering, access control, and test coverage.
```

## Limitations

- Provides design guidance and examples, but does not generate a complete production integration automatically.
- Does not replace provider documentation for current SDK methods or model support.
- Requires environment-specific testing for authentication, permissions, timeouts, and external service behavior.
- Does not evaluate deployed tools or verify their runtime security.

## Best Practices

- Describe when each tool applies, what every parameter means, and what the result contains.
- Validate all model-provided inputs before execution and return errors that support recovery.
- Test tool selection and argument quality with the target model alongside conventional unit and integration tests.

## Anti Patterns

- Do not expose many overlapping tools with vague names or descriptions.
- Do not return silent failures, generic errors, or unstructured implementation details.
- Do not concatenate user input into SQL, shell commands, URLs, or other executable contexts.

## Security Audit

- Audited at: 2026-07-23T20:28:54.01\+00:00
- Summary: All seven static findings are false positives caused by documentation syntax, illustrative API calls, or ordinary prose. The skill contains guidance and examples, with no executable scripts or prompt injection evidence.

## Stats

- Views: 104
- Downloads: 7
- Favorites: 0
- Popularity score: 0