I'm Smars, a data engineer in Toronto. I work in banking, where an AI agent is only useful if you can prove what it did and stop it from doing what it shouldn't. That is the problem I build for.
A harness that wraps GitHub Copilot in VS Code so a team can let an agent work on real repositories.
- Guardrails Rules enforced by code. Hooks check every tool call before it runs. Out-of-scope writes,
.gitchanges and dangerous commands are rejected, whatever the prompt says. - Tasks that survive a new session. Each task is a directory holding the requirement, plan, progress and result, so a fresh session picks up where the last one stopped.
- Task Observability Dashboard. User can get the statistic information of every task, etc time consumed, tool calling, or token costs etc.
- Multi-agents collaboration Wrap up the most useful multi-agents scenarios only (Independent reviewer, Map & Reduce, AutoRun etc...)
- Human-approved write-back. The agent works in a task directory and reads the target repository's
main. Changes reach the target only through a promote you approve, on a new branch. - Context Engineering Team can integrate their customized knowledge-base with our harness framework, then agents will be able to do your team's work.
Python standard library only, no pip install. Runs on macOS and Windows.
| Project | What it is | |
|---|---|---|
| EComDWH-BatchDataProcessingPlatform | A five-layer data warehouse for e-commerce orders on a self-managed cluster: Spark, Airflow, ClickHouse, with CI/CD and monitoring. |  |
| azure-cloud-datapipeline-EDA | A lakehouse pipeline on Azure Databricks with Delta Lake, Unity Catalog and Data Factory. |  |
Coding agents
Harness engineering
Context engineering
Multi-agent workflow
Evals and verification
Languages
Data processing and lakehouse
Orchestration, quality and governance
Cloud and DevOps
