/shipmates-deslop

Discovery grades every finding safe, review, architectural or protected, then files the work as an epic with sub-issues. Nothing is edited: you review the breakdown and choose what ships.

1. 1

   Survey

   Bound the scope, fix the mode, and inventory which analysers the repo already has — nothing is scanned until the scope is named.

   The run does this itself — no specialist.

2. 2

   Census

   Inventory every finding — including files sitting in the wrong layer for the repo's own stated architecture — then grade each by the risk of touching it.

   Always

   sdet security-engineer technical-writer

   Also sit when

   • architectnear-duplicate and structural classes
3. 3

   Weigh

   Near-duplicates get a semantic diff and over-built machinery gets a state-versus-consumer count, so nothing is filed as an impression.

   Crew

   architect

4. 4

   Map

   Every risky finding is traced back to the features that depend on it — and a missing feature can only raise a grade, never lower one.

   Crew

   product-manager site-reliability-engineer performance-engineer

5. 5

   File

   Findings become an epic with sub-issues, grouped one reviewable diff and one owner at a time, each carrying its evidence and its acceptance criteria.

   The run files these itself; the tracker is the only thing it writes.

6. 6

   Choose

   You review the epic, confirm the breakdown, then take the whole thing, a single sub-issue, or a bundle — and it hands off to the workflow that ships.

   Your gate. Nothing proceeds until you say so.

Shared machinery — how commands work: worktrees, the CI gate, the review board, and merge modes.