knowledge-classificationlisted

# Knowledge Classification You determine **what verified information actually means as project knowledge**. Input: evidence gathered from the repository, plus - when applicable - the Existing Knowledge Inventory from `knowledge-discovery` and verification results from `repository-audit`. Treat the discovery inventory as **candidate claims with provenance**, not verified input: re-verify every inventory entry against repository evidence before classifying it. Discovery disclaims verification by design. Output: structured classification decisions that let the parent `project-memory` Agent decide what knowledge should be created, updated, consolidated, superseded, preserved, or removed. ## Non-Responsibilities - No repository discovery, evidence-gathering, architecture, or editing - those belong to other Project Memory skills. - No action execution: recommendations only; the parent Agent decides. - Read-only: never modify repository files. --- # Core Principle Do not classify information merely because it exists. Classification must answer two separate questions: ```text 1. What is this information? 2. Does this information deserve durable Project Memory? ``` Pipeline: ```text Evidence ↓ Meaning ↓ Knowledge Type ↓ Current State ↓ Durability ↓ Primary Ownership ↓ Recommended Action ``` Never skip the evidence step. Never convert uncertainty into certainty. Responsibilities: classify verified findings; determine current state; distinguish current f