netgrade-digital · realM1lF · Jun 4, 2026
diff --git a/README.md b/README.md
@@ -24,7 +24,7 @@
 
 ## Why this exists
 
-Teams use many AI tools (Cursor, Claude Code, Zed, Codex, …). Each session starts cold: no shared workflows, no institutional memory, and knowledge trapped in chats or private notes.
+Teams use many AI tools (Cursor, Claude Code, Zed, Codex, Kimi, …). Each session starts cold: no shared workflows, no institutional memory, and knowledge trapped in chats or private notes.
 
 **Shared Agents** gives you one place to:
 
@@ -92,7 +92,7 @@ After pulling Core + team (ff-only):
 
 1. **Skill symlinks** — `skills/` + `team/skills/` → `~/.agents/skills`, `~/.claude/skills`, …
 2. **Rule symlinks** — Cursor: `rules/` + `team/rules/` → `~/.cursor/rules/`
-3. **Team-rules blocks** — Zed, Codex, Claude Code, Gemini, Windsurf, …: `<!-- shared-agents:team-rules:begin/end -->` in each tool's `AGENTS.md` / `CLAUDE.md`
+3. **Team-rules blocks** — Zed, Codex, Claude Code, Gemini, Windsurf, Kimi (best-effort), …: `<!-- shared-agents:team-rules:begin/end -->` in each tool's `AGENTS.md` / `CLAUDE.md`
 
 Also runs quietly via IDE session hooks where configured. **`sa install`** is still required once per tool (hooks, base `<!-- shared-agents:begin -->` block).
 
@@ -138,7 +138,7 @@ Install path defaults to `~/.shared-agents`. Team data lives under `team/` insid
 |-------|--------|
 | Learnings workflow | [docs/learnings.md](docs/learnings.md) |
 | Paths agents must use | [docs/canonical-paths.md](docs/canonical-paths.md) |
-| Per-tool adapters | [adapters/](adapters/) (Cursor, Claude Code, Zed, …) |
+| Per-tool adapters | [adapters/](adapters/) (Cursor, Claude Code, Zed, Kimi, …) |
 | Team skills & rules CLI | Skill [sa-cli](skills/sa-cli/SKILL.md) · `sa skill` · `sa rule` |
 | Migrating legacy `learnings/` | [docs/migration-team-data.md](docs/migration-team-data.md) |
 | CLI reference (full) | Skill `sa-cli` in [skills/sa-cli/SKILL.md](skills/sa-cli/SKILL.md) |

diff --git a/adapters/README.md b/adapters/README.md
@@ -15,6 +15,7 @@ Per-tool wiring is defined in [`manifest.json`](manifest.json). **`sa install`**
 | [Copilot](copilot/README.md) | `~/.copilot/AGENTS.md` | Marker block |
 | [Aider](aider/README.md) | `~/.aider/AGENTS.md` | Marker block |
 | [OpenClaw](openclaw/README.md) | Entrypoint wrapper (headless) | Read paths under `$SHARED_AGENTS_HOME` |
+| [Kimi Code CLI](kimi/README.md) | `~/.kimi/AGENTS.md` | Marker block (best-effort) |
 | [Generic](generic/README.md) | Copy-paste instructions | Manual |
 
 Team content (all tools): `sa skill new|list|rm` · `sa rule new|list|rm` — see root [README.md](../README.md).

diff --git a/adapters/generic/README.md b/adapters/generic/README.md
@@ -23,6 +23,7 @@ Many agents follow the **AGENTS.md pattern**:
 | OpenCode | `~/.config/opencode/AGENTS.md` |
 | Claude Code | `~/.claude/CLAUDE.md` or project `CLAUDE.md` |
 | Gemini CLI | `~/.gemini/GEMINI.md` |
+| Kimi Code CLI | `~/.kimi/AGENTS.md` (best-effort — not auto-loaded) |
 | Project-level | `./AGENTS.md`, `./CLAUDE.md`, `.cursorrules` |
 
 ## Add first-class support

diff --git a/adapters/kimi/README.md b/adapters/kimi/README.md
@@ -0,0 +1,52 @@
+# Kimi Code CLI
+
+> **Best-effort workaround — not a full integration.**
+>
+> Kimi does **not** natively load a global instructions file at session start (unlike Claude Code or Cursor). `sa install` creates `~/.kimi/AGENTS.md` anyway, but you must copy its contents into your prompt manually or use a wrapper script. `sa sync` keeps the file up to date so you always have the latest block ready to paste.
+
+Auto-configured by **`sa install`** when `~/.kimi` exists or the `kimi` binary is on `PATH`.
+
+Re-run **`sa sync`** after team changes (skills + rules) · First-time: **`sa install`** · Status: **`sa check`**
+
+## Global instructions
+
+Merged into **`~/.kimi/AGENTS.md`** (applies to all projects):
+
+| Marker block | Content |
+|--------------|---------|
+| `<!-- shared-agents:begin/end -->` | Sync + learnings workflow |
+| `<!-- shared-agents:team-rules:begin/end -->` | Core + team rules from `$SHARED_AGENTS_HOME/rules/` and `team/rules/*.mdc` |
+
+Your own sections **above** those markers (e.g. communication style) are never overwritten. The team-rules block is **refreshed on every `sa sync`** for detected tools.
+
+## How to use the AGENTS.md content
+
+Because Kimi does not read `~/.kimi/AGENTS.md` automatically, pick one of these workflows:
+
+1. **Copy-paste** — open `~/.kimi/AGENTS.md` and paste the `<!-- shared-agents:begin/end -->` block into your first prompt at the start of each session.
+2. **Wrapper script** — create a small shell alias or script that reads `~/.kimi/AGENTS.md` and passes it as context to `kimi`.
+3. **Project-level** — copy the block into a project `AGENTS.md` or `.kimi/AGENTS.md` if your project setup supports it.
+
+## Team skills & rules
+
+| Type | Team path | On Kimi |
+|------|-----------|---------|
+| Skills | `team/skills/<name>/SKILL.md` | Not auto-loaded — copy manually from `~/.agents/skills/` if needed |
+| Rules | `team/rules/<slug>.mdc` | Merged into `~/.kimi/AGENTS.md` team-rules block |
+
+```bash
+sa skill new | sa skill list | sa skill rm [name]
+sa rule new  | sa rule list  | sa rule rm [slug]
+```
+
+Wizards commit/push by default (`--no-git` to skip). Teammates run **`sa sync`**. Optional rule frontmatter: `targets: [kimi, cursor]` — omit for all adapters.
+
+## Sync
+
+No built-in session hook — the AGENTS.md block instructs the user to run `sync.sh pull` as the first command, but Kimi does not execute it automatically. You must trigger `sa sync` manually or copy the updated block before each session.
+
+## Why this is marked "best-effort"
+
+- No native session hook (Cursor `hooks.json`, Claude Code `settings.json`).
+- No native global instructions loader (Zed, Codex, Aider load `AGENTS.md` automatically).
+- The `~/.kimi/` directory is **not** an official Kimi config path; it is a convention chosen by this project so that `sa install` / `sa sync` have a stable target file.
diff --git a/adapters/manifest.json b/adapters/manifest.json
@@ -129,6 +129,15 @@
       "name": "Any AGENTS.md / CLAUDE.md CLI",
       "detect": null,
       "docs": "adapters/generic/README.md"
+    },
+    {
+      "id": "kimi",
+      "name": "Kimi Code CLI",
+      "detect": "~/.kimi",
+      "detect_bins": ["kimi"],
+      "docs": "adapters/kimi/README.md",
+      "agents_md": "~/.kimi/AGENTS.md",
+      "note": "Best-effort workaround — Kimi has no native global instructions loader. AGENTS.md is not read automatically; copy content manually or use a wrapper."
     }
   ]
 }
diff --git a/docs/canonical-paths.md b/docs/canonical-paths.md
@@ -101,7 +101,7 @@ Bei **`sa sync`** (täglich) bzw. **`sa install`** (Erst-Setup):
 | Adapter-Typ | Installation |
 |-------------|--------------|
 | **Cursor** | Symlinks → `~/.cursor/rules/*.mdc` |
-| **AGENTS.md / CLAUDE.md** (Zed, Codex, **Claude Code**, Gemini, Windsurf, …) | Marker-Block `<!-- shared-agents:team-rules:begin/end -->` |
+| **AGENTS.md / CLAUDE.md** (Zed, Codex, **Claude Code**, Gemini, Windsurf, Kimi best-effort, …) | Marker-Block `<!-- shared-agents:team-rules:begin/end -->` |
 | **Eigene Datei am Ziel** (kein Symlink, Cursor) | **Nicht überschrieben** |
 
 Quellen: `$SHARED_AGENTS_HOME/rules/` (Core) + `$SHARED_AGENTS_HOME/team/rules/` (Team).

diff --git a/skills/sa-cli/SKILL.md b/skills/sa-cli/SKILL.md
@@ -231,7 +231,7 @@ Team rules: `$SHARED_AGENTS_HOME/team/rules/*.mdc` (flat). Core: `$SHARED_AGENTS
 After edit: commit/push team repo → teammates **`sa sync`** (links skills + rules automatically).
 
 - **Cursor:** symlinks → `~/.cursor/rules/` (local non-symlink files preserved)
-- **Zed, Codex, Claude Code (`~/.claude/CLAUDE.md`), Gemini, Windsurf, …:** merged `<!-- shared-agents:team-rules:begin/end -->` block in each tool's agents file
+- **Zed, Codex, Claude Code (`~/.claude/CLAUDE.md`), Gemini, Windsurf, Kimi (best-effort), …:** merged `<!-- shared-agents:team-rules:begin/end -->` block in each tool's agents file
 
 Optional frontmatter: `targets: [zed, claude-code]` — omit for all adapters.