discoverlisted

# Discover (phase 1 of 7) Ground the task in **this codebase** and settle the reuse question before any ideation. This phase was split out of Brainstorm (decision reversal, v0.4): exploration + reuse-scan are *discovery*, not *ideation* — folding them into Brainstorm over-fit it and killed creative breadth. Discover does the grounding; Brainstorm (phase 2) then ideates freely on top of it. Runs **inline on the main thread** (it owns the state write; a forked subagent cannot spawn subagents). ## Iron Law ``` NO NEW CLASS, SERVICE, UTILITY, CONFIG, OR ENDPOINT BEFORE A REUSE SCAN ``` The `PreToolUse` write gate enforces this: until `reuse_scan=true` (recorded here), every production write is denied — in **every** complexity tier. Discover is the one phase the fast lane never skips. ## The decision ladder (what the scan decides) The reuse-scan is not only "does the project already have it?" — it answers the full lazy-senior-dev ladder for each thing the task would build, **stopping at the first rung that fits** (create-time depth: `skills/implement/references/minimalism.md`): ``` 0. need-to-exist? → no: DROP it (YAGNI) drop 1. JDK / Java stdlib does it? ┐ 2. Spring / installed starter? ├ → use it, write nothing framework 3. already-declared dependency?┘ (check build.gradle/pom.xml's classpath) 4. existing PROJECT code? → adopt as-is | extend it (cite file:line) adopt | extend 5. nothing fits