surveying-codebaselisted

# Surveying-Codebase — Cold-start the memory of an onboarded project **Announce at the start (in pt-br):** "Usando surveying-codebase para popular a memória do projeto a partir da própria codebase." **Input:** the codebase of the project being onboarded — no run, no `findings.md`. **This is the cold-start twin of `distilling-learnings`.** Both route durable knowledge by blast-radius into the repo-committed memory model (see `CLAUDE-HARNESS-MEMORY-MODEL.md`). The ONLY difference is the SOURCE: distilling reads a run's transient `findings.md`; surveying reads the **codebase itself**. The routing machine — durability test, blast-radius classes, destinations, retire-on-promote — is identical and is **reused, never re-implemented**. There is no `learnings.md`. Pairs with `initializing-projects`: init creates the STRUCTURE (nested `CLAUDE.md` stubs + root router table); surveying fills the CONTENT. --- ## When to use - A legacy/existing project enters the harness with no accumulated memory. - `orchestrating-delivery` Phase 0 finds `.claude/memory/MEMORY.md` and nested `CLAUDE.md` empty on a non-trivial project. - Manually via `/surveying-codebase`. - **Re-run** when the codebase changes substantially (new subsystem, big refactor). **One-time / on-demand — not a per-run step.** When NOT to use: a fresh/empty project (nothing durable to extract yet) → just run `initializing-projects`. --- ## Pipeline 1. **Confirm structure** — the root router table and nested `CLAUDE.md` s