Human-in-the-loop pattern as working code: the model drafts, a person approves or overrides, only approved items execute, every step lands in an audit log. State machine enforced, tested offline.

agentsaudit-log

claudehuman-in-the-loop

python

claude-ops-toolkit

Repo · 0 stars · Python

▲ 0

The discipline I run Claude Code with, generalized: planning gates with checkable acceptance criteria, a hypothesis-driven debugging loop, pre-ship checks, and a report-only QC reviewer agent.

agent-skillsclaude-codedeveloper-tools

python

school-finance-agent

Repo · 0 stars · Python

▲ 0

Claude tool-use agent over the school-finance MCP server, with a 12-question graded eval suite (contains, number_close, tool_used checks). Domain tools to agent to evals, end to end.

agents

claudeevals

mcp

python

school-finance-mcp

Repo · 0 stars · Python

▲ 0

MCP server for California school finance: SACS account-string validation/decoding and school-meal reimbursement, built with the official MCP Python SDK (FastMCP).

california

claudefastmcp

mcp

python

Stack

Vercel

Claude Code

Suno

3 tools

About

Operations leader who ships production systems solo, then deploys them with the team that runs them. Full-stack and agentic systems on Claude. Founder, Databros