Public Resources from SevenBelow — Compliance OS
Public showcase repo for SevenBelow work: production Claude Code framework, AI-engineering bootcamp solutions, and legacy GSD tooling. Each top-level directory ships its own README with deeper detail.
Production-grade Claude Code framework — v1. Foundation of the SevenBelow Compliance OS MVP.
Highlights:
- Three-repo governance pattern — Authority (
compliance-os-goldmaster-prds), Orchestration (sevenbelow-orchestrator), Execution (compliance-core,compliance-ui) - GSD (Get Shit Done) protocol — disciplined plan → execute → verify loop
- CARL (Context-Aware Rule Layer) — governance rules layered across repos
- Claude Skills system — composable, reusable skill modules
- Operating contracts — strict CLAUDE.md per repo enforcing scope and boundaries
global/— shared settings, hooks, agents, commands across all v1 repos
Includes the original mentor shoutout (Charles Dove / CC Strategic AI, James Goldbach / Agent Architects) and the full architecture deep-dive with mermaid diagrams.
Status: v1 — superseded by an in-progress v2 architecture. Published as the reference implementation it grew out of.
DataExpert.io 2026 AI Engineering Bootcamp homework solutions. Each subdir has its own README; original problem statements and solutions are not reproduced.
Highlights:
claude-code-homework/— Release Radar: GitHub-activity tool (issue triage, PR summaries, commit digests, stakeholder emails) built as Claude Code skills + CLI batch mode, with a guardrail chain (PII redaction, citation, schema validation, uncertainty/insufficient-context handling)langchain-homework-2026/— three LangChain/LangGraph assignments: context-overflow management, bounded infinite-researcher loop, MCP middleware. Includes per-assignment runtime logs, writeups, and an aggregate token-report toolopenclaw-week-3-homework/— token-efficiency optimization on an OpenClaw Alfred workflow. Self-hosted Langfuse for trace capture, before/after metrics extraction, evidence collected on both live and simulated runsopenclaw-week-4-homework/— debug-and-harden on a broken OpenClaw agent. Packagedopenclaw_agentmodule, LangSmith evals, safety/tool/model/classifier test suites, before/after metrics, preflight check
GSD Checkpoint Skill — historical showcase. Lightweight state-preservation skill for Claude Code sessions running the GSD protocol. Captures session state to prevent context loss across compaction events.
Status: superseded by GSD v2. Kept here as a reference for what a self-contained, installable Claude Code skill looks like end-to-end (
SKILL.mdbody, install script, state-verification helper).
| If you want… | Go to |
|---|---|
| Production Claude Code framework + GSD/CARL deep dive | claude-setups/compliance-os-project-v1/ |
| Bootcamp solutions (Claude Code, LangChain, OpenClaw) | dataexpertio-2026-AI-Eng-Bootcamp/ |
| Example installable Claude Code skill | gsd-skills/checkpoint/ |
See LICENSE.