@fde-lab/configuration-engineer

FDE Lab

FDE Lab is a collection of 11 AI-native Forward Deployed Engineer (FDE) workers. These workers handle various complex infrastructure and software tasks deterministically and output machine-readable JSON.

Try it

npx @fde-lab/incident-engineer

AI Consumption Contract

Every worker supports standard AI-native execution modes and a predictable contract designed specifically for external orchestrators.

1. Discover worker: Use --manifest to discover worker capabilities, input scenarios, and expected side-effects.

   npx @fde-lab/incident-engineer --manifest

2. Understand invocation: The manifest (manifest.inputs and manifest.examples) documents how to execute the worker.

3. Understand result: The manifest (manifest.outputs.WorkerResult) provides a strict JSON Schema detailing exactly what data the worker will return.

4. Execute: Invoke the worker via npx, passing required inputs.

5. Parse: Use --json to retrieve the strictly-structured output. The stdout is guaranteed to contain only the clean JSON result envelope.

   npx @fde-lab/incident-engineer --json > result.json

6. Interpret: The JSON payload provides actionable context including status, confidence, next_steps, and artifacts.

7. Use process exit code: The Node.js process propagates the granular Python semantic exit codes (e.g., 5 for Inconclusive, 0 for Success) as documented in manifest.exit_codes. This allows orchestrators to make immediate fallback or circuit-breaker decisions without parsing stdout.

Workers

Architecture

AI / Orchestrator
        ↓
   --manifest
        ↓
    FDE Worker
        ↓
   --json result
        ↓
AI / Orchestrator

Repository structure

• packages/: Individual worker npm packages
• fde_lab/: Shared Python runtime and POC implementations
• registry/: Capability and package registries
• docs/: POC documentation

npm / GitHub

The npm packages wrap the Python workers for easy distribution. The packages contain pre-bundled Python environments that install transparently into a local cache directory on first run.

Source: https://github.com/smg99/fde-lab