Demonstrates a practical, organized approach to building AI agent workspaces with layered memory and reusable skills, suitable for live demos and training.
One app anchors every demo in the two-hour session. Meta-repo layout — start AI sessions at the repo root; that's where the hot memory and skills are auto-loaded:
Demo/ ← start AI sessions HERE (the workspace hub root)
├── AGENTS.md hot memory: constitution, routing, conventions
├── CLAUDE.md minimal pointer → AGENTS.md
├── mars.yaml catalog of project repos (multi-repo status/sync)
├── .claude/
│ └── skills/ implement-feature · route-to-flow · release-note
│ git-commit · git-open-pr · coding-convention
├── agent-hub/ the rest of the team brain
│ ├── knowledge-base/ cold memory: reference-spec · user-guide · decision-note
│ ├── flow-registry/ "what code implements this?"
│ └── tickets/ DEMO-101 (the Part 3 live-demo ticket)
├── taskflow-api/ FastAPI task tracker — homepage shows the session agenda
└── scan-simulator/ lightweight mock of the heavy external scan service
cd scan-simulator && pip install -r requirements.txt && uvicorn app.main:app --port 9000
cd taskflow-api && pip install -r requirements.txt && uvicorn app.main:app --port 8000
Open http://127.0.0.1:8000 — agenda + live task board.
| Session moment | What to do |
|---|---|
| Minute 0 | Open the homepage — agenda lives in the app |
| Part 2: contrast | Same request, plain prompt vs. agent started at repo root |
| Part 2: skills | Invoke release-note on recent tickets/commits |
| Part 3: memory layers | Walk the layers — AGENTS.md, .claude/skills/, agent-hub/knowledge-base/ |
| Part 3: force multipliers | route-to-flow on a question; show the simulator answering /v1/scan |
| Part 3: live pipeline | Run implement-feature on ticket DEMO-101, tests vs. simulator, board updates live |
Similar projects matched by category, topics, and programming language.
Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.