← Files Data Engineering CopilotARCHIVED FILE

README.md

2.09 KB · Nov 9, 2026 · 00:48 UTC

↓ Download file

See the change to this file →

# Data Engineering Copilot — v0.2.0

Tekris Labs skills-only plugin for production-focused data architecture, pipeline implementation, debugging, incident recovery and validation.

## What's new
- 18 focused skills (instead of one overlarge instruction file).
- Original skill retained in `docs/original-v0.1.0-skill.md`, all seven original technical references preserved in `references/original/`.
- New coverage for Microsoft Fabric/OneLake, dbt and SQL warehouse transforms, lineage/quality/security, incident response, execution evidence and real-subagent coordination.
- Reusable architecture, incident, backfill, model, quality, release and agent-handoff templates.
- 24 **unexecuted prompt-level** benchmarks and a small offline fixture checker with unit tests.
- Four Tekris Labs website/support/legal URLs and three starter prompts.

## Routing
Start with `skills/data-engineering/SKILL.md`. Specialized skills under `skills/<name>/SKILL.md` trigger only for their actual tasks. Reference docs are loaded on demand.

## Runtime truth
This ZIP is **skills-only**, has no embedded MCP endpoint, and cannot by itself inspect live Databricks/Airflow/Fabric/warehouse systems, execute production jobs, or spawn agent workers. Those operations require actual authorized host tools, user permissions and a suitable environment. A written SQL/PySpark example is not executed code. Never claim successful tests without results.

## Quality validation
From plugin root:

```sh
python3 scripts/validate_package.py
python3 -m unittest discover -s tests -v
python3 scripts/reconcile_fixture.py benchmarks/fixtures/reconciliation.json
```

The fixture checker runs locally on example JSON only. It is not a production correctness certificate and does not evaluate the real plugin in ChatGPT. To evaluate product behavior, execute and score the cases in `benchmarks/scenarios.json` against a supported host.

## Submission
Use the same published package identity `production-data-engineering-copilot`. The OpenAI Platform update flow chooses the owned existing plugin rather than creating a duplicate. See `SUBMISSION.md` and `UPGRADE_AUDIT.md`.

SHA-256: c32dcc889681dad4a01d16cb677486e403231dc50bb681149bc65b76969e394d