Mirror of ~/.claude/projects/-Users-pp-bot/memory/ as of 2026-04-27. Files: - MEMORY.md (index) - dasheng-codebase.md (Gitea + IM family + 取经团 worker layout) - feedback_bot_node_field.md (bots.json node field is informational) Per-user Claude memory is Mac-local. This repo is the canonical backup so a fresh machine / EC2 / colleague workstation can prime new Claude Code windows with the same project knowledge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
1.5 KiB
1.5 KiB
name, description, type, originSessionId
| name | description | type | originSessionId |
|---|---|---|---|
| bots.json node field is informational, not routing | When working with the dasheng claude-worker-universe bot registry, do not block planning or PRDs on resolving the node field for sub-bots; many run on colleague machines that the user does not own. | feedback | c2c569ed-e20b-4000-a286-8d48afea8d86 |
In jx-im/claude-worker-universe/config/bots.json, the node field on each bot entry is informational, not behavioral. Many sub-bots run on colleague machines outside the user's control. Do not treat unknown / unverified node values as blockers.
Why: User feedback 2026-04-25 iter2 of /ralph: "子机器人很多都是同事的机器 所以不要纠结在哪里" — many sub-bots are colleague-owned. The dispatch system uses Gitea labels + dashboard KV for routing, not the node field; bots self-register at startup.
How to apply:
- When designing or PRD-ing changes that touch bots.json, do NOT add acceptance criteria like "every bot has a verified node value" — accept
node="unknown"or stale values. - When asking the user about a new bot, DO NOT make the node a blocker question; default to
MacBook(this user's primary machine) and let the colleague update it later if they own the bot. - For 文殊菩萨 specifically: it runs on the same node as 唐僧 (MacBook) per separate instruction; that's fixed.
- Routing decisions (H5 vs iOS, qa_priority verify-first) come from
platforms+qa_priority+role_typefields — never fromnode.