pattern-retrospectivelisted

# Pattern-retrospective workflow You're producing a study / blog post / handoff / spec / retrospective that draws conclusions from Bryce's past Claude sessions. Apply this rigor before publishing. The chat-arch v2 Rev 3 plan describes the substrate that will eventually automate this. Until that ships (the chat-arch repo's `_planning/chat-arch-v2-rev3-plan.md`), you are the curator, falsifier, and substrate. The outputs you write become the seed data when chat-arch's automation lands — write them so they're machine-readable later. ## §1 — The load-bearing meta-discipline **Before specifying any external system's requirements, audit that system first.** If you're writing a "handoff to X" / "spec for X" / "improvements to X" / "what X should do" document, spend 30 minutes reading X's code, locked spec, and CLAUDE.md before writing. Lead the document with a "Context: what already exists in X" section. If you can't write that section, you can't write the document. This applies recursively. When you cite a fact about an external system, verify it the same way you'd verify any other agent's cited claim. "X already does Y" is a falsifiable claim — falsify it before relying on it. **Why this rule exists:** the May 2026 handoff to chat-arch failed this check. Roughly 60% of its proposed "chat-arch requirements" already existed in the chat-arch codebase — Phase 1's log-location discovery, Phase 3's false-positive filter, Phase 6's external verification, Phase 7's aggregate metric