investigatelisted

# investigate — turn a live symptom into grounded 🔲 Backlog tasks `/investigate` is the **upstream of `/groom`**. Grooming brings `🔲 Backlog` → `🗂️ Ready`; this produces the grounded Backlog. It takes an operator-reported symptom ("sessions are erroring after the last deploy", "the design launch keeps failing") and turns it into **`🔲 Backlog` tasks rooted in the live operational record** — not in what the code looks like it should do. **Read-only diagnosis is the default posture.** Observe the record, root-cause it, file. Diagnosis never mutates another session's git/PR/worktree to learn what it can read. An interactive run can escalate to operator-authorized action (fire a verify/ops session, correct a status/Type, an operator-granted DB write) along the ladder in **§ The mutation boundary** — never improvised, never across the two forbidden lines (managed PRs; session worktrees/git). The skill's value is a verification discipline that refuses to trust premises, statuses, or the checkout HEAD, plus a repeatable **investigate → classify → file** loop. Follow it even when the symptom looks like a one-line fix. > **Not `/ops`'s Investigation mode.** `/ops` works a **pre-filed `🔎` Investigation task** already on > a milestone board, under its journal/staging machinery. `/investigate` is **ad-hoc and > operator-pointed**: there is no task yet — the symptom is the input, filed Backlog tasks are the > output. It shares `/ops`'s by-value / falsify / root-cause disciplines b