fde-auditlisted

# @fde-audit ## Token efficiency Load `context.md` if it exists -- otherwise start cold. Do not load other `.fde/` files until the audit produces them. The point of this skill is to establish ground truth, not assume it. ## Purpose You're not starting fresh. Someone else was here before you, or the team has been building something for weeks. Before touching anything, you need to know what's real, what's assumed, and what's load-bearing. ## Opening You're picking up someone else's engagement. Let them unload — what actually works, what's theater, what's held together with duct tape. Don't interrupt; you're separating fact from story. ## What to listen for **What works vs what's assumed to work:** systems nobody has actually tested end-to-end. **Load-bearing workarounds:** scripts, manual steps, or hacks that became critical infrastructure. Touching these without knowing breaks everything. **What the previous person left behind:** any `.fde/` context, docs, notes, Slack history. Read it all before forming any opinion. **The single highest risk right now:** what's the thing that, if it breaks today, stops the customer's business? **Who knows what:** which team member holds knowledge that exists nowhere else. ## One follow-up (if needed) What's the one thing they'd be insane to touch blind? That's usually the load-bearing wall. ## What to produce A clear audit that tells any FDE, including one who just arrived, exactly where things stand: - What exists and what act