# Build eve Agents with Version-Matched Documentation

Agent development can fail when guidance does not match the installed eve version. This skill directs your assistant to bundled documentation before project changes.

## Install

```bash
npx skillstore add vercel/eve
```

## Metadata

- Status: approved
- Slug: vercel-eve
- Skillstore revision: r1
- Version status: missing
- Tree hash: 33acfda3d07bd958399a6c20201d9b127b1bd69e706673aa9d744521f5940b26
- Author: vercel
- GitHub username: vercel
- License: MIT
- Repository: https://github.com/vercel/eve/tree/77874cd4c51cec475d67dc213dcdc5c314c82f11/skills/eve
- Ref: 726d2ed3a01be16d942ed15f4801cc68bec3995a
- 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: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/vercel-eve
- Manifest: https://skillstore.pages.dev/api/skills/vercel-eve/manifest

## Capabilities

- Directs the assistant to documentation bundled in node\_modules/eve/docs/.
- Identifies the bundled README as the documentation index and recommended reading order.
- Requires reading relevant guides before writing eve code.
- Provides npm installation and npx agent scaffolding instructions when eve is absent.
- Defines its scope around eve project creation, editing, and debugging across agent configuration areas.

## Use Cases

- Start an eve Agent Project: Locate setup guidance and scaffolding instructions, then consult bundled documentation before creating an agent.
- Maintain Agent Configuration: Find version-matched guides before changing tools, connections, channels, subagents, or schedules.
- Investigate Agent Behavior: Use the documentation index to locate relevant guidance before debugging an eve project or reviewing its evaluations.

## Prompt Templates

### Locate the Documentation

```
Check whether eve is installed. If available, read its bundled documentation index and identify the beginner guides. Do not install anything.
```

### Prepare a New Agent

```
Help prepare a new eve agent named support-agent. Read available setup documentation first. Ask before installing dependencies or executing scaffolding commands.
```

### Update a Schedule

```
Review my eve project and read the installed version's scheduling guide. Propose the requested schedule change and explain how to verify it.
```

### Investigate Tool Failures

```
Investigate recurring tool failures in my eve agent. Read relevant bundled guides, inspect authorized project files, and propose focused changes with verification steps.
```

## Limitations

- The skill contains instructions only; it includes no implementation scripts, reference guides, or runnable examples.
- Detailed development guidance depends on documentation available in the installed eve package.
- Installing or scaffolding requires npm tooling and may require network access and permission to execute package code.
- The skill does not independently implement, test, or guarantee agent behavior.

## Best Practices

- Read the installed package's documentation index and relevant guides before making changes.
- Review package identity and request authorization before installation or CLI execution.
- Keep changes focused and verify behavior using checks documented for the installed version.

## Anti Patterns

- Treating this short skill as a complete eve API reference.
- Using remembered APIs without checking documentation for the installed version.
- Executing installation or scaffolding commands without reviewing permissions and package trust.

## Security Audit

- Audited at: 2026-10-05T16:43:59.385\+00:00
- Summary: All five static findings are false positives caused by Markdown inline code and fenced blocks, not executable shell backticks. The installation commands are ordinary framework setup instructions, but execution requires trust in the npm package. No evidence found of prompt injection, credential theft, or data exfiltration in the reviewed skill.

## Stats

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