Remove non-project governance files from git tracking

- Deleted: KIMI.md, GEMINI.md, AGENTS.md, CLAUDE.md, docs/SOP.md
- Updated .gitignore to prevent re-tracking

这些文件是从 kimi-prj 复制过来的 Cat Cafe 治理配置，不属于 Lan-manager 项目本身

[宪宪/glm-5 🐾]

.gitignore

@@ -5,5 +5,12 @@
 .gemini/
 .kimi/
 
+# Agent governance files (not project-specific)
+KIMI.md
+GEMINI.md
+AGENTS.md
+CLAUDE.md
+docs/SOP.md
+
 # System files
 .DS_Store

AGENTS.md

@@ -1,30 +0,0 @@
-<!-- CAT-CAFE-GOVERNANCE-START -->
-> Pack version: 1.3.0 | Provider: codex
-
-## Cat Cafe Governance Rules (Auto-managed)
-
-### Hard Constraints (immutable)
-- **Public local defaults**: use frontend 3003 and API 3004 to avoid colliding with another local runtime.
-- **Redis port 6399** is Cat Cafe's production Redis. Never connect to it from external projects. Use 6398 for dev/test.
-- **No self-review**: The same individual cannot review their own code. Cross-family review preferred.
-- **Identity is constant**: Never impersonate another cat. Identity is a hard constraint.
-
-### Collaboration Standards
-- A2A handoff uses five-tuple: What / Why / Tradeoff / Open Questions / Next Action
-- Vision Guardian: Read original requirements before starting. AC completion ≠ feature complete.
-- Review flow: quality-gate → request-review → receive-review → merge-gate
-- Skills are available via symlinked cat-cafe-skills/ — load the relevant skill before each workflow step
-- Shared rules: See cat-cafe-skills/refs/shared-rules.md for full collaboration contract
-
-### Quality Discipline (overrides "try simplest approach first")
-- **Bug: find root cause before fixing**. No guess-and-patch. Steps: reproduce → logs → call chain → confirm root cause → fix
-- **Uncertain direction: stop → search → ask → confirm → then act**. Never "just try it first"
-- **"Done" requires evidence** (tests pass / screenshot / logs). Bug fix = red test first, then green
-
-### Knowledge Engineering
-- Documents use YAML frontmatter (feature_ids, topics, doc_kind, created)
-- Three-layer info architecture: CLAUDE.md (≤100 lines) → Skills (on-demand) → refs/
-- Backlog: BACKLOG.md (hot) → Feature files (warm) → raw docs (cold)
-- Feature lifecycle: kickoff → discussion → implementation → review → completion
-- SOP: See docs/SOP.md for the 6-step workflow
-<!-- CAT-CAFE-GOVERNANCE-END -->

CLAUDE.md

@@ -1,30 +0,0 @@
-<!-- CAT-CAFE-GOVERNANCE-START -->
-> Pack version: 1.3.0 | Provider: claude
-
-## Cat Cafe Governance Rules (Auto-managed)
-
-### Hard Constraints (immutable)
-- **Public local defaults**: use frontend 3003 and API 3004 to avoid colliding with another local runtime.
-- **Redis port 6399** is Cat Cafe's production Redis. Never connect to it from external projects. Use 6398 for dev/test.
-- **No self-review**: The same individual cannot review their own code. Cross-family review preferred.
-- **Identity is constant**: Never impersonate another cat. Identity is a hard constraint.
-
-### Collaboration Standards
-- A2A handoff uses five-tuple: What / Why / Tradeoff / Open Questions / Next Action
-- Vision Guardian: Read original requirements before starting. AC completion ≠ feature complete.
-- Review flow: quality-gate → request-review → receive-review → merge-gate
-- Skills are available via symlinked cat-cafe-skills/ — load the relevant skill before each workflow step
-- Shared rules: See cat-cafe-skills/refs/shared-rules.md for full collaboration contract
-
-### Quality Discipline (overrides "try simplest approach first")
-- **Bug: find root cause before fixing**. No guess-and-patch. Steps: reproduce → logs → call chain → confirm root cause → fix
-- **Uncertain direction: stop → search → ask → confirm → then act**. Never "just try it first"
-- **"Done" requires evidence** (tests pass / screenshot / logs). Bug fix = red test first, then green
-
-### Knowledge Engineering
-- Documents use YAML frontmatter (feature_ids, topics, doc_kind, created)
-- Three-layer info architecture: CLAUDE.md (≤100 lines) → Skills (on-demand) → refs/
-- Backlog: BACKLOG.md (hot) → Feature files (warm) → raw docs (cold)
-- Feature lifecycle: kickoff → discussion → implementation → review → completion
-- SOP: See docs/SOP.md for the 6-step workflow
-<!-- CAT-CAFE-GOVERNANCE-END -->

GEMINI.md

@@ -1,30 +0,0 @@
-<!-- CAT-CAFE-GOVERNANCE-START -->
-> Pack version: 1.3.0 | Provider: gemini
-
-## Cat Cafe Governance Rules (Auto-managed)
-
-### Hard Constraints (immutable)
-- **Public local defaults**: use frontend 3003 and API 3004 to avoid colliding with another local runtime.
-- **Redis port 6399** is Cat Cafe's production Redis. Never connect to it from external projects. Use 6398 for dev/test.
-- **No self-review**: The same individual cannot review their own code. Cross-family review preferred.
-- **Identity is constant**: Never impersonate another cat. Identity is a hard constraint.
-
-### Collaboration Standards
-- A2A handoff uses five-tuple: What / Why / Tradeoff / Open Questions / Next Action
-- Vision Guardian: Read original requirements before starting. AC completion ≠ feature complete.
-- Review flow: quality-gate → request-review → receive-review → merge-gate
-- Skills are available via symlinked cat-cafe-skills/ — load the relevant skill before each workflow step
-- Shared rules: See cat-cafe-skills/refs/shared-rules.md for full collaboration contract
-
-### Quality Discipline (overrides "try simplest approach first")
-- **Bug: find root cause before fixing**. No guess-and-patch. Steps: reproduce → logs → call chain → confirm root cause → fix
-- **Uncertain direction: stop → search → ask → confirm → then act**. Never "just try it first"
-- **"Done" requires evidence** (tests pass / screenshot / logs). Bug fix = red test first, then green
-
-### Knowledge Engineering
-- Documents use YAML frontmatter (feature_ids, topics, doc_kind, created)
-- Three-layer info architecture: CLAUDE.md (≤100 lines) → Skills (on-demand) → refs/
-- Backlog: BACKLOG.md (hot) → Feature files (warm) → raw docs (cold)
-- Feature lifecycle: kickoff → discussion → implementation → review → completion
-- SOP: See docs/SOP.md for the 6-step workflow
-<!-- CAT-CAFE-GOVERNANCE-END -->

KIMI.md

@@ -1,30 +0,0 @@
-<!-- CAT-CAFE-GOVERNANCE-START -->
-> Pack version: 1.3.0 | Provider: kimi
-
-## Cat Cafe Governance Rules (Auto-managed)
-
-### Hard Constraints (immutable)
-- **Public local defaults**: use frontend 3003 and API 3004 to avoid colliding with another local runtime.
-- **Redis port 6399** is Cat Cafe's production Redis. Never connect to it from external projects. Use 6398 for dev/test.
-- **No self-review**: The same individual cannot review their own code. Cross-family review preferred.
-- **Identity is constant**: Never impersonate another cat. Identity is a hard constraint.
-
-### Collaboration Standards
-- A2A handoff uses five-tuple: What / Why / Tradeoff / Open Questions / Next Action
-- Vision Guardian: Read original requirements before starting. AC completion ≠ feature complete.
-- Review flow: quality-gate → request-review → receive-review → merge-gate
-- Skills are available via symlinked cat-cafe-skills/ — load the relevant skill before each workflow step
-- Shared rules: See cat-cafe-skills/refs/shared-rules.md for full collaboration contract
-
-### Quality Discipline (overrides "try simplest approach first")
-- **Bug: find root cause before fixing**. No guess-and-patch. Steps: reproduce → logs → call chain → confirm root cause → fix
-- **Uncertain direction: stop → search → ask → confirm → then act**. Never "just try it first"
-- **"Done" requires evidence** (tests pass / screenshot / logs). Bug fix = red test first, then green
-
-### Knowledge Engineering
-- Documents use YAML frontmatter (feature_ids, topics, doc_kind, created)
-- Three-layer info architecture: CLAUDE.md (≤100 lines) → Skills (on-demand) → refs/
-- Backlog: BACKLOG.md (hot) → Feature files (warm) → raw docs (cold)
-- Feature lifecycle: kickoff → discussion → implementation → review → completion
-- SOP: See docs/SOP.md for the 6-step workflow
-<!-- CAT-CAFE-GOVERNANCE-END -->

docs/SOP.md

@@ -1,24 +0,0 @@
----
-topics: [sop, workflow]
-doc_kind: note
-created: 2026-04-19
----
-
-# Standard Operating Procedure
-
-## Workflow (6 steps)
-
-| Step | What | Skill |
-|------|------|-------|
-| 1 | Create worktree | `worktree` |
-| 2 | Self-check (spec compliance) | `quality-gate` |
-| 3 | Peer review | `request-review` / `receive-review` |
-| 4 | Merge gate | `merge-gate` |
-| 5 | PR + cloud review | (merge-gate handles) |
-| 6 | Merge + cleanup | (SOP steps) |
-
-## Code Quality
-
-- Biome: `pnpm check` / `pnpm check:fix`
-- Types: `pnpm lint`
-- File limits: 200 lines warn / 350 hard cap