auditflow

Think before you build. Audit after you ship.

An [Agent Skill](https://agentskills.io) for thinking about systems before (and after) you build them.

Tell your AI agent to trace a flow — it walks through your codebase recording every step in SQLite, then exports Mermaid diagrams, Markdown reports, JSON, and YAML. Works with [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [OpenAI Codex CLI](https://github.com/openai/codex), and any agent that supports the [Agent Skills spec](https://agentskills.io/specification).

Why

Most teams jump straight to code. The architecture lives in someone's head, a stale Confluence page, or a whiteboard photo rotting in Slack.

audit-flow makes system thinking a first-class artifact:

• Before building — sketch the flow as a DAG. Which layers? What triggers what? Where does data move? Your brainstorm session produces a queryable database and exportable diagrams, not ephemeral notes.
• After building — trace the real implementation. Compare what you planned vs what you built. The ideation flow and the documentation flow live in the same database.
• When things break — trace the bug path. Your incident post-mortem links back to the original design flow. Findings accumulate across sessions.
• Over time — 10 audits later, you have a queryable map of your entire system. New engineer joins? audit.py list shows every flow ever traced.

It's TDD for architecture. Trace the flow f