Local-first nutrition MCP for agents. Intake writes stay gated. Prefer MCP tools if connected; otherwise the package CLI. Use when the user wants Nourish data or actions through an agent.
Installs into .claude/skills of the current project.
Are you the author of Skill?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/davidmosiah-skill-wellness-nourish)
---
name: nourish
description: >
Local-first nutrition MCP for agents. Intake writes stay gated. Prefer MCP tools if connected; otherwise the package CLI.
Use when the user wants Nourish data or actions through an agent.
---
# Nourish — skill or MCP
Same binary either way. Do not duplicate the API client.
## Choose a surface
**MCP** — tools appear natively after stdio/HTTP config:
```json
{ "mcpServers": { "nourish": { "command": "npx", "args": ["-y", "wellness-nourish"] } } }
```
Do not put mutation flags in that snippet.
**Skill / CLI** — no MCP client required. Same tools:
```bash
npx -y wellness-nourish call nourish_connection_status --json '{}'
```
If MCP tools named `nourish_*` are already available, use them. Do not also shell out.
## Loop
1. Call `nourish_connection_status` (or `doctor --json` when that exists).
2. Use read tools as asked.
3. Stop on `USER_ACTION_REQUIRED`. Do not invent env flags. Do not enable mutations from this skill.
## Never
- Paste tokens into git, chat logs, or the prompt
- Copy a mutations-enabled assignment into config
