Skills & Connectors

Skills

A skill is a prompt instruction stored as a SKILL.md file:

---
name: fast-analytics-query
description: Query FAST for brand/category/retailer analytics
connectors:
  - fast-query
memory_reads:
  - fast:last_response
memory_writes:
  - fast:last_query
---

# Fast Analytics Query

## Instructions

When a user asks about brand performance, category trends, or retailer data,
use the fast_query tool to send their question to FAST...

Skills tell the LLM what to do. Connectors tell it how.

Connectors

A connector is a deterministic function with typed config:

tools/fast-query/
  TOOL.md         # Metadata
  impl.py         # Async Python function
  config.yaml     # Environment-aware config + secrets schema

Implementation

async def fast_query(app, trigger: str) -> dict:
    from mesh.config.secrets import resolve_connector_config
    config = resolve_connector_config("fast-query")
    # config = {"base_url": "https://stage-...", "user_id": "...", ...}

Config Schema

schema:
  base_url:
    type: url
    required: true
  user_id:
    type: secret
    secret_key: fast-chat-api/user-id

environments:
  dev:
    base_url: https://stage-fast-chat-api.../v1/chat/sync
  prod:
    base_url: https://prod-fast-chat-api.../v1/chat/sync

Non-secret values per environment in config.yaml. Secrets from AWS Secrets Manager or .secrets.yaml (local dev).

Creating

# Create a skill
mesh skills create my-skill --description "What it does"

# Create a connector
mesh tools create my-tool --function-name my_func --tag api