@hayasaka7/haya-pet 0.3.2 → 0.3.3

package/CHANGELOG.md

@@ -7,6 +7,15 @@ All notable changes to HAYA Pet are documented here. This project adheres to
 > 0.2.0 npm publish; they are listed under 0.2.1, which is the first version that
 > ships them.
 
+## [0.3.3]
+
+### Fixed
+- **Claude Code subagent completion no longer changes the main session status.**
+  Claude Code can emit `SubagentStop` after the main agent has already stopped,
+  so treating that event as `idle` could make the pet react to a stale subagent
+  completion instead of the main agent's real state. The Claude hook adapter now
+  ignores `SubagentStop`; the main turn still ends on Claude's `Stop` event.
+
 ## [0.3.2]
 
 ### Changed

package/README.md

@@ -64,7 +64,7 @@ npm install -g @hayasaka7/haya-pet
 From source:
 
 ```bash
-git clone <repo-url> haya-pet
+git clone https://github.com/HAYASAKA7/HAYA-PET.git haya-pet
 cd haya-pet
 npm install
 npm link

package/docs/known-issues.md

@@ -2,6 +2,20 @@
 
 Issues found in live use, with their current status.
 
+## ✅ Resolved: Claude Code subagent completion changed the main session status
+
+- **Symptom:** In Claude Code multi-agent runs, the main agent could already be
+  stopped while a subagent was still finishing. When that late subagent emitted
+  `SubagentStop`, the pet treated it as a main-session `idle` update and could
+  show a misleading working/done transition after the main agent had settled.
+- **Root cause:** The Claude hook table mapped `SubagentStop` to `idle`. That is
+  only safe if subagent completion is ordered before the main turn finishes, which
+  Claude Code does not guarantee.
+- **Fix:** Claude `SubagentStop` is now ignored. Main-session idle still comes
+  from Claude's real `Stop` hook, while late subagent completion cannot override
+  the current main-agent state. Codex keeps its separate behavior because Codex
+  uses `Stop` as the only idle signal and treats `SubagentStop` as mid-turn.
+
 ## ✅ Resolved: false "waiting for approval" while Codex auto-reviews an approval (Approve for me)
 
 - **Symptom:** Running Codex under the pet with the **"Approve for me"** preset
@@ -228,8 +242,9 @@ observation (`--observe`) or L1 lifecycle as the fallback. Current state:
   lifecycle status). Live in-session status is **opt-in** via `HAYA_PET_HOOKS=1`,
   which injects a settings file (`claude --settings <stable-file>`, no change to
   your global config) wiring Claude's `UserPromptSubmit`/`PreToolUse`/`PostToolUse`/
-  `Notification`/`PreCompact`/`Stop`/`SubagentStop` events to `haya-pet state
-  <state>`, reported to the daemon over the IPC pipe. `PreToolUse` distinguishes
+  `Notification`/`PreCompact`/`Stop` events to `haya-pet state <state>`, reported
+  to the daemon over the IPC pipe. `SubagentStop` is intentionally ignored because
+  it is not a main-turn idle signal. `PreToolUse` distinguishes
   file-editing tools (`Edit`/`Write`/`MultiEdit`/`NotebookEdit` → *editing files*)
   from other tools (→ *running tools*) via the hook `matcher`. **Why opt-in:**
   injecting hooks makes Claude show a one-time *review hooks* trust prompt; the

package/docs/troubleshooting.md

@@ -16,6 +16,7 @@ deferred problems with known root causes.
 | Terminal scroll / Shift+Tab / backspace odd while a CLI runs under `haya-pet run` | Fixed — `haya-pet run` now uses native passthrough by default (full fidelity). If you opted into `--observe`, drop it. See [known-issues.md](known-issues.md). |
 | Pet shows only **idle/lifecycle** while **Claude Code** works | Live in-session status is opt-in: run `haya-pet hooks on` once (persisted). The first `haya-pet run` afterward shows a one-time Claude *review hooks* prompt — approve it. Also make sure the companion is running (`haya-pet start`). Check the toggle with `haya-pet hooks status`. |
 | Typing doesn't work / **Claude Code** TUI frozen under `haya-pet run` | You have hooks enabled and Claude is showing its *review hooks* trust prompt (approve it once), or your Claude is too old for `--settings`. Run `haya-pet hooks off` (or set `HAYA_PET_NO_HOOKS=1`) for native passthrough with lifecycle-only status — typing and Shift+Tab work normally. |
+| Pet changes status after a **Claude Code** subagent finishes, even though the main agent already stopped | Fixed — Claude `SubagentStop` is ignored because it is not a reliable main-turn state. Update to the latest version and restart the wrapped Claude session so the new hook settings are used. |
 | Pet shows only **idle/lifecycle** while **Codex** works | Live status is opt-in: run `haya-pet hooks on` once (persisted, global), then `haya-pet run --client codex -- codex`; approve Codex's one-time *review hooks* prompt. `thinking`/`idle` come from hooks, `running_tool`/`editing_files` from a transcript watcher, and approval states from the `PermissionRequest` hook plus a guardian-review watcher. |
 | Pet showed **waiting for approval** while **Codex** auto-reviewed the request ("Approve for me") | Fixed — with `approvals_reviewer = auto_review` (legacy `guardian_subagent`) Codex's guardian decides without asking you; the pet now shows **reviewing** during the assessment, then **working** on an allow verdict or **thinking** on a deny. *Waiting for approval* still shows when Codex actually asks you (`approvals_reviewer = "user"`). |
 | **Codex** live status didn't turn on / you pass your own `-p`/`--profile` | Codex allows only one profile, so haya-pet skips hook injection when you supply your own and prints a notice. Drop your `-p` for that run to get live status, or accept lifecycle-only. |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hayasaka7/haya-pet",
-  "version": "0.3.2",
+  "version": "0.3.3",
   "type": "module",
   "description": "Generic AI CLI pet runtime foundation.",
   "keywords": [

package/packages/adapters/src/claude-hooks.js

@@ -34,8 +34,7 @@ const HOOK_TABLE = Object.freeze([
   { event: "PermissionDenied", state: "idle", summary: "denied" },
   { event: "PreCompact", state: "compacting" },
   { event: "Stop", state: "idle" },
-  { event: "StopFailure", state: "idle", summary: "stopped" },
-  { event: "SubagentStop", state: "idle" }
+  { event: "StopFailure", state: "idle", summary: "stopped" }
 ]);
 
 // Resolve the pet state for a Claude event. `detail` is the tool name for

package/packages/adapters/test/claude-hooks.test.mjs

@@ -9,12 +9,15 @@ test("mapClaudeEventToState covers activity events", () => {
   assert.equal(mapClaudeEventToState("PreCompact"), "compacting");
   assert.equal(mapClaudeEventToState("Stop"), "idle");
   assert.equal(mapClaudeEventToState("StopFailure"), "idle");
-  assert.equal(mapClaudeEventToState("SubagentStop"), "idle");
   assert.equal(mapClaudeEventToState("PermissionDenied"), "idle");
   assert.equal(mapClaudeEventToState("PermissionRequest"), "waiting_approval");
   assert.equal(mapClaudeEventToState("Unknown"), undefined);
 });
 
+test("mapClaudeEventToState ignores Claude SubagentStop", () => {
+  assert.equal(mapClaudeEventToState("SubagentStop"), undefined);
+});
+
 test("mapClaudeEventToState branches PreToolUse on tool name", () => {
   assert.equal(mapClaudeEventToState("PreToolUse", "Bash"), "running_tool");
   assert.equal(mapClaudeEventToState("PreToolUse", "Edit"), "editing_files");
@@ -80,8 +83,9 @@ test("buildClaudeHookSettings includes all subscribed events", () => {
   for (const event of [
     "UserPromptSubmit", "PreToolUse", "PostToolUse", "PostToolUseFailure",
     "PermissionRequest", "Notification", "PermissionDenied", "PreCompact",
-    "Stop", "StopFailure", "SubagentStop"
+    "Stop", "StopFailure"
   ]) {
     assert.ok(settings.hooks[event], `missing hook event ${event}`);
   }
+  assert.equal(settings.hooks.SubagentStop, undefined);
 });