trustmaplisted

# Trustmap Auto mode routes every tool call through a classifier that blocks anything aimed outside your environment. By default it trusts exactly two things: the working directory, and the current repo's configured remotes. Everything else — your other GitHub org, the LAN device you flash builds to, your internal API — reads as a potential exfiltration target and gets blocked. `autoMode.environment` is the prose block that tells the classifier what's actually yours. Claude Code ships a wizard that drafts that block by scanning your machine when you accept it. **On a fresh install that wizard is the better tool** — it reads the machine directly and needs no history, while this skill has nothing to read. Once a corpus exists the trade reverses, and answering from extracted events differs in three ways that matter: - **Usage weighting.** A repo under `$HOME` is a candidate; a repo you pushed to twenty-seven times is infrastructure. The wizard's own output labels its repo list "CANDIDATES, not vetted context." Here the hit counts do that work. - **Reach.** Codex sessions and backfilled git history are in this corpus and outside a Claude-Code-only scan. - **Re-runnability.** Every proposal is diffed against what's already in your settings, so the second run proposes only the delta. The wizard is one-shot; it won't notice the internal host you started using last month. Run it on first setup, and again when a project stops being routine — a new org, a new deploy targe