# Build and Configure AI Agents in Cargo

Configuring agents requires coordinating prompts, models, knowledge, tools, and releases. This skill guides Cargo configuration through documented CLI workflows and examples.

## Install

```bash
npx skillstore add getcargohq/cargo-ai
```

## Metadata

- Status: approved
- Slug: getcargohq-cargo-ai
- Version: 2.4.0
- Author version: 2.4.0
- Skillstore revision: r1
- Version status: valid
- Tree hash: c2503e866a6b1cb4dff2081bff1a5b515bfa000466bb74c7edb3868f3c71287f
- Author: getcargohq
- GitHub username: getcargohq
- License: MIT
- Repository: https://github.com/getcargohq/cargo-skills/tree/92c825ad4ba1bbbbeff96ae85e62187279f3854d/cargo-ai
- Ref: 85f27aab6597ed2f415c40f9093c67022ab251a7
- 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, filesystem, network
- Quality score: 79
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/getcargohq-cargo-ai
- Manifest: https://skillstore.pages.dev/api/skills/getcargohq-cargo-ai/manifest

## Capabilities

- Create, inspect, rename, organize, and remove Cargo agents using documented CLI commands.
- Discover agent templates and adapt their prompts, model settings, and temperature.
- Configure draft releases and deploy versioned agent configurations.
- Attach existing knowledge files or libraries through release resources for retrieval.
- Publish curated MCP servers, connect external MCP clients, and disable selected tools.
- Inspect, update, and remove memories within agent, user, or workspace scopes.

## Use Cases

- Configure a Lead Research Agent: Adapt a research template, select model settings, and prepare an agent release for prospect research.
- Attach an Existing Knowledge Library: Inspect an agent draft and attach approved knowledge resources before deploying the updated configuration.
- Expose Selected Workspace Tools: Publish a curated MCP server with selected actions and read-only model resources for a coding assistant.

## Prompt Templates

### Inspect Available Agents

```
Confirm my active Cargo workspace, then list agents and templates. Summarize available configurations without changing anything.
```

### Prepare an Agent Draft

```
Find a company classification template and prepare a new agent draft. Use low temperature and show the configuration before deployment.
```

### Attach Approved Knowledge

```
Attach my approved existing library to the specified agent draft. Preserve existing resources and tools, summarize changes, and request approval before deployment.
```

### Publish a Curated MCP Server

```
Plan a curated Cargo MCP server for these approved tools and model columns. Keep data read-only and require approval before publishing.
```

## Limitations

- Requires the external Cargo CLI, authentication, a workspace, and sufficient permissions.
- Knowledge uploads and library management belong to cargo-content; agent execution belongs to cargo-orchestration.
- Structured output and heartbeat settings require direct API calls because documented CLI flags are unavailable.
- Model availability and command support depend on your Cargo environment and CLI version.

## Best Practices

- Confirm the workspace and target UUIDs before writes, deletions, or deployments.
- Read current configuration and preserve complete arrays when updates replace actions, resources, or capabilities.
- Use trusted MCP endpoints, minimal tool access, and read-only data resources unless writes are explicitly required.

## Anti Patterns

- Deploying a draft without reviewing its prompt, model, resources, and tools.
- Assuming knowledge uploads automatically attach resources to an agent.
- Treating published workspace MCP servers and consumed external MCP clients as interchangeable.

## Security Audit

- Audited at: 2026-10-12T17:15:10.141\+00:00
- Summary: All 167 static alerts are false positives involving Markdown formatting, documentation links, Cargo resource identifiers, and documented service or example URLs. No evidence found of malicious prompt injection, credential exfiltration, or unauthorized host reconnaissance. This review covers the supplied documentation, not the external CLI package or live Cargo and MCP services.

## Stats

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