@gotgenes/pi-subagents 10.1.0 → 10.2.1

package/CHANGELOG.md

@@ -5,6 +5,35 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [10.2.1](https://github.com/gotgenes/pi-packages/compare/pi-subagents-v10.2.0...pi-subagents-v10.2.1) (2026-05-27)
+
+
+### Documentation
+
+* **pi-subagents:** renumber Phase 15 steps to match execution order ([598bb65](https://github.com/gotgenes/pi-packages/commit/598bb653ac8b63756e8c00dfcf19d3167e2dbc37))
+* **pi-subagents:** revise Phase 15 roadmap for Agent-born-complete vision ([e04583e](https://github.com/gotgenes/pi-packages/commit/e04583e75bfc1314674a6f3181762a26733fb830))
+* plan push exec/registry relay deps to runner construction ([#231](https://github.com/gotgenes/pi-packages/issues/231)) ([646b4d5](https://github.com/gotgenes/pi-packages/commit/646b4d5085e0f7d36a397b43b3b46e0537c3141f))
+* **retro:** add planning stage notes for issue [#231](https://github.com/gotgenes/pi-packages/issues/231) ([dc0daee](https://github.com/gotgenes/pi-packages/commit/dc0daee634c17cf2a40336e27f551bfa2ce0e249))
+* **retro:** add retro notes for issue [#228](https://github.com/gotgenes/pi-packages/issues/228) ([d5b563b](https://github.com/gotgenes/pi-packages/commit/d5b563b6484cbd6a89cd7e9e87ebd431aed128fc))
+* **retro:** add TDD stage notes for issue [#231](https://github.com/gotgenes/pi-packages/issues/231) ([28094ae](https://github.com/gotgenes/pi-packages/commit/28094ae812141ea1c93a22be50ed29d31b7a979a))
+* update architecture for runner self-contained ([#231](https://github.com/gotgenes/pi-packages/issues/231)) ([80dd339](https://github.com/gotgenes/pi-packages/commit/80dd339d7dee9b312b52af2b74756c5748619a49))
+
+## [10.2.0](https://github.com/gotgenes/pi-packages/compare/pi-subagents-v10.1.0...pi-subagents-v10.2.0) (2026-05-27)
+
+
+### Features
+
+* **pi-subagents:** add run lifecycle methods to Agent ([2a378f1](https://github.com/gotgenes/pi-packages/commit/2a378f1c82e977bdfee25931ab449757e364d589))
+
+
+### Documentation
+
+* **pi-subagents:** update architecture for async startAgent ([941eb10](https://github.com/gotgenes/pi-packages/commit/941eb109e71e4c51d5bb37a2a46ffc12f618d949))
+* plan async startAgent and RunHandle dissolution ([#228](https://github.com/gotgenes/pi-packages/issues/228)) ([647adf8](https://github.com/gotgenes/pi-packages/commit/647adf853fec63ea53afd63bc8204c89a6194bbe))
+* **retro:** add planning stage notes for issue [#228](https://github.com/gotgenes/pi-packages/issues/228) ([8dd9f8a](https://github.com/gotgenes/pi-packages/commit/8dd9f8ab7082c08e424b1b4a9557253af2ce584b))
+* **retro:** add retro notes for issue [#227](https://github.com/gotgenes/pi-packages/issues/227) ([78a4d64](https://github.com/gotgenes/pi-packages/commit/78a4d645f524465c64bf0b6ba1bcca37858e8721))
+* **retro:** add TDD stage notes for issue [#228](https://github.com/gotgenes/pi-packages/issues/228) ([ab497c5](https://github.com/gotgenes/pi-packages/commit/ab497c57723666d0635a0a08f9eecc06576da549))
+
 ## [10.1.0](https://github.com/gotgenes/pi-packages/compare/pi-subagents-v10.0.1...pi-subagents-v10.1.0) (2026-05-27)
 
 

package/docs/architecture/architecture.md

@@ -55,7 +55,7 @@ flowchart TB
         direction TB
         AgentManager["AgentManager<br/>(spawn, queue, abort)"]
         AgentRunner["agent-runner<br/>(session, turns, results)"]
-        AgentRecord["Agent<br/>(status, behavior: abort/steer/worktree)"]
+        Agent["Agent<br/>(status, behavior: abort/steer/worktree/run lifecycle)"]
         ParentSnapshot["ParentSnapshot<br/>(frozen parent state)"]
         Worktree["worktree<br/>(git isolation)"]
     end
@@ -101,7 +101,7 @@ flowchart TB
 
 ```mermaid
 classDiagram
-    class AgentRecord {
+    class Agent {
         +id: string
         +type: SubagentType
         +description: string
@@ -124,6 +124,12 @@ classDiagram
         +queueSteer(message)
         +flushPendingSteers(session)
         +setupWorktree(worktrees, isolation)
+        +completeRun(result, worktrees)
+        +failRun(err, worktrees)
+        +wireSignal(signal, onAbort)
+        +attachObserver(unsub)
+        +releaseListeners()
+        +setOnRunFinished(fn)
     }
 
     class AgentManager {
@@ -160,7 +166,7 @@ classDiagram
         +hasRunning(): boolean
     }
 
-    AgentManager --> AgentRecord : creates/manages
+    AgentManager --> Agent : creates/manages
     AgentManager --> ParentSnapshot : receives at spawn
     SubagentsService --> AgentManager : wraps via adapter
     AgentManager --> AgentTypeRegistry : resolves types
@@ -266,7 +272,6 @@ src/
 │   ├── parent-snapshot.ts          immutable spawn-time parent state
 │   ├── execution-state.ts          session/output phase state
 │   ├── permission-bridge.ts        optional bridge to pi-permission-system registry
-│   ├── run-handle.ts               per-run cleanup lifecycle
 │   ├── worktree.ts                 git worktree isolation
 │   ├── worktree-state.ts           worktree phase state
 │   └── usage.ts                    token usage tracking
@@ -594,23 +599,21 @@ export interface ParentSessionInfo {
 
 `AgentSpawnConfig` now carries `parentSession?: ParentSessionInfo` instead of three flat optional fields.
 
-#### RunOptions (12 fields → extract RunContext) — done ([#169][169])
+#### RunOptions (12 fields → extract RunContext) — done ([#169][169]), updated by [#231]
 
-The `RunOptions` bag mixes execution parameters with context information.
-`RunContext` was extracted and nested as `RunOptions.context`:
+`RunContext` was extracted and nested as `RunOptions.context` in #169.
+Issue #231 moved the two static dependencies (`exec`, `registry`) to `RunnerDeps` on `ConcreteAgentRunner`, leaving `RunContext` with only per-call fields:
 
 ```typescript
-/** Parent execution context — where/who is running. */
+/** Per-call execution context — fields that vary per spawn. */
 export interface RunContext {
-  exec: ShellExec;
-  registry: AgentConfigLookup;
   cwd?: string;
   parentSession?: ParentSessionInfo;
 }
 ```
 
 The remaining `RunOptions` fields (`model`, `maxTurns`, `signal`, `isolated`, `thinkingLevel`, `defaultMaxTurns`, `graceTurns`, `onSessionCreated`) are genuine execution parameters.
-`RunOptions` now has 9 fields: 1 nested `context: RunContext` plus 8 flat execution fields.
+`RunOptions` now has 9 fields: 1 nested `context: RunContext` (2 per-call fields) plus 8 flat execution fields.
 
 #### SessionConfig (11 fields → extract ToolFilterConfig) — done ([#168][168])
 
@@ -681,24 +684,66 @@ See [phase-14-strip-policy.md](history/phase-14-strip-policy.md) for details.
 
 ## Improvement roadmap (Phase 15 — domain model evolution)
 
-Phase 15 addresses the anemic domain model in the lifecycle layer.
-`AgentRecord` is a data bag — identity, status transitions, and stats — but no behavior.
-`AgentManager` reaches into records 37 times, doing work that belongs on the agent.
-Per-agent state (pending steers, abort logic, run lifecycle) is scattered across the manager, `RunHandle`, and a manager-level Map.
+Phase 15 evolves `Agent` from a passive state machine into an object that **owns its entire execution lifecycle**.
+
+Steps 1–2 (complete) moved per-agent behavior from `AgentManager` onto `Agent`: abort, steer buffering, worktree setup, and run lifecycle methods (`completeRun`, `failRun`).
+However, Agent still cannot *run itself*.
+`AgentManager.startAgent()` orchestrates the entire execution: calling the runner, handling session creation, wiring observers, and cleaning up worktrees.
+The manager reaches into Agent 10 times across `spawn()` + `startAgent()` — writing to `notification`, `execution`, and `promise` after construction, passing its own `worktrees` and `runner` as method arguments, and threading `onSessionCreated` callbacks through three layers.
+
+The remaining steps address this by making **Agent born complete**: constructed with all dependencies and configuration, owning its entire execution lifecycle.
+
+### Architecture target
+
+Agent receives three concerns at construction:
+
+| Concern     | Fields                                                                        | Lifetime                  |
+| ----------- | ----------------------------------------------------------------------------- | ------------------------- |
+| Identity    | id, type, description, invocation                                             | Immutable                 |
+| Run config  | snapshot, prompt, model, isolation, maxTurns, thinking, signal, parentSession | Immutable per-run         |
+| Shared deps | runner, worktrees                                                             | Shared service references |
+
+`Agent.run()` encapsulates the full execution lifecycle:
+
+1. Set up worktree internally (knows its own isolation mode, has worktrees).
+2. Call `this.runner.run()` (has the runner).
+3. Handle session creation internally: set `execution`, flush pending steers, attach record-observer.
+4. Notify lifecycle observer (started, session created, completed, compacted).
+5. Clean up worktree on completion or error.
+6. Transition status.
+
+`AgentManager` becomes a collection manager + concurrency controller:
+
+- Creates complete Agent objects, stores them in the map.
+- Decides when to run (immediate or queue) and calls `agent.run()`.
+- Provides high-level actions: abort, list, cleanup.
+- Does *not* own the runner, worktrees, or any run-orchestration logic.
+
+The queue stores agent IDs, not `SpawnArgs`.
+When capacity opens, the manager looks up the agent and calls `agent.run()` — the agent already has everything.
+
+The `onSessionCreated` callback that currently threads through `AgentSpawnConfig` → `startAgent` → `RunOptions` → runner disappears.
+Agent handles session creation internally during `run()` and notifies external observers via the lifecycle observer pattern.
+
+The synchronous-throw contract for worktree failure (introduced in Step 2's hoist) is replaced by a uniform async error surface.
+Worktree failures inside `agent.run()` propagate through the promise.
+For background agents, errors surface via `get_subagent_result` and appear in `/agents`.
+For foreground agents, `spawnAndWait` awaits the promise naturally.
 
 The scheduling concern (queue, concurrency counter, drain) is tangled into `AgentManager` alongside collection management and run orchestration.
 `notifyConcurrencyChanged()` is a scheduling method exposed as a public API so settings can poke the queue — a cross-concern leak.
 
 ### Findings summary
 
-| Finding                                                       | Category     | Impact | Risk | Priority |
-| ------------------------------------------------------------- | ------------ | ------ | ---- | -------- |
-| `AgentRecord` is anemic — no behavior, manager reaches in 37× | B: Oversized | 5      | 3    | 15       |
-| Scheduling tangled into `AgentManager` (3 fields, 3 methods)  | A: Coupling  | 4      | 2    | 12       |
-| `startAgent` uses `.then()`/`.catch()` instead of async/await | C: Callbacks | 3      | 2    | 10       |
-| `onSessionCreated` callback flows through 3 layers            | C: Callbacks | 3      | 2    | 10       |
-| `resume()` duplicates observer subscribe/unsubscribe pattern  | A: Redundant | 2      | 1    | 8        |
-| `exec`/`registry` relay-only deps on `AgentManager`           | C: Coupling  | 2      | 1    | 6        |
+| Finding                                                            | Category     | Impact | Risk | Priority |
+| ------------------------------------------------------------------ | ------------ | ------ | ---- | -------- |
+| ~~`AgentRecord` is anemic — no behavior, manager reaches in 37×~~  | B: Oversized | 5      | 3    | ✅       |
+| Agent cannot run itself — manager orchestrates 10 external touches | C: Coupling  | 5      | 3    | 15       |
+| Scheduling tangled into `AgentManager` (3 fields, 3 methods)       | A: Coupling  | 4      | 2    | 12       |
+| ~~`startAgent` uses `.then()`/`.catch()` instead of async/await~~  | C: Callbacks | 3      | 2    | ✅       |
+| ~~`onSessionCreated` callback flows through 3 layers~~             | C: Callbacks | 3      | 2    | subsumed |
+| `resume()` duplicates observer subscribe/unsubscribe pattern       | A: Redundant | 2      | 1    | 8        |
+| `exec`/`registry` relay-only deps on `AgentManager`                | C: Coupling  | 2      | 1    | 6        |
 
 ### Step 1: Evolve AgentRecord into Agent with behavior — [#227] ✅ Complete
 
@@ -713,53 +758,66 @@ Move per-agent behavior from `AgentManager` into the agent:
 - Smell: B (anemic domain model) + C (manager reaching into records)
 - Outcome: `AgentManager` delegates via Tell-Don't-Ask; per-agent state lives on the agent
 
-### Step 2: Convert startAgent to async/await — [#228]
+### Step 2: Convert startAgent to async/await — [#228] ✅ Complete
 
-Convert `startAgent` from synchronous (returns void, assigns `record.promise` to a `.then()`/`.catch()` chain) to `async` (returns `Promise<void>`, uses try/catch).
+Converted `startAgent` to `async` with `try/catch` and dissolved `RunHandle` into `Agent` methods.
 `spawn()` assigns `record.promise = this.startAgent(...)` instead of calling `startAgent()` synchronously.
+`Agent` gained run lifecycle methods: `completeRun`, `failRun`, `wireSignal`, `attachObserver`, `releaseListeners`, `setOnRunFinished`.
+Worktree setup was hoisted to callers (`spawn`, `drainQueue`) to preserve the synchronous-throw contract.
 
 - Depends on: #227
-- Target: `src/lifecycle/agent-manager.ts`
+- Target: `src/lifecycle/agent-manager.ts`, `src/lifecycle/agent.ts`
 - Smell: C (raw promise callbacks)
-- Outcome: zero `.then()`/`.catch()` in `agent-manager.ts`
+- Outcome: zero `.then()`/`.catch()` in `agent-manager.ts`; `RunHandle` deleted; Agent owns run lifecycle
 
-### Step 3: Replace onSessionCreated callback with observer method — [#229]
+### Step 3: Push exec/registry relay deps to runner construction — [#231] ✅
 
-Add `onSessionCreated(agent, session)` to `AgentManagerObserver`.
-Remove the `onSessionCreated` callback from `AgentSpawnConfig`.
-Tool-layer code subscribes via the observer pattern instead of passing callbacks through the spawn config.
+`exec` and `registry` moved from `AgentManager` to `ConcreteAgentRunner` via a new `RunnerDeps` interface.
+`RunContext` shrunk from 4 to 2 per-call fields (`cwd`, `parentSession`).
+`AgentManagerOptions` shrunk from 7 to 5 fields.
 
-- Target: `src/lifecycle/agent-manager.ts`, `src/tools/background-spawner.ts`, `src/tools/foreground-runner.ts`
-- Smell: C (callback flowing through 3 layers)
-- Outcome: `AgentSpawnConfig` loses one callback field; session notification uses the observer pattern
+- Target: `src/lifecycle/agent-manager.ts`, `src/lifecycle/agent-runner.ts`, `src/index.ts`
+- Smell: C (relay-only dependencies)
+- Outcome: `AgentManager` loses 2 fields; `AgentManagerOptions` shrinks from 7 to 5 fields; runner is self-contained
+
+### Step 4: Agent born complete — Agent.run() absorbs startAgent — [#229]
+
+Agent receives `runner`, `worktrees`, and a lifecycle observer at construction.
+Agent creates its own `NotificationState` from `parentSession.toolCallId` — no external write.
+`Agent.run()` encapsulates the entire execution lifecycle: worktree setup, runner invocation, session-creation handling, observer wiring, worktree cleanup, and status transitions.
+`startAgent` is deleted from `AgentManager`.
+The `onSessionCreated` callback is removed from `AgentSpawnConfig` — Agent handles session creation internally and notifies via the lifecycle observer.
+`SpawnArgs` is deleted — Agent has its config from construction.
+
+`AgentManager.spawn()` becomes: create complete Agent, put in map, call `agent.run()` or queue the agent ID.
+
+- Depends on: #228, #231
+- Target: `src/lifecycle/agent.ts`, `src/lifecycle/agent-manager.ts`, `src/tools/background-spawner.ts`, `src/tools/foreground-runner.ts`
+- Smell: C (manager orchestrates 10 external touches on Agent) + C (callback flowing through 3 layers)
+- Outcome: Agent owns its entire execution lifecycle; `startAgent`, `SpawnArgs`, `onSessionCreated` callback deleted; zero post-construction writes from `AgentManager`
 
-### Step 4: Extract ConcurrencyQueue from AgentManager — [#230]
+### Step 5: Extract ConcurrencyQueue from AgentManager — [#230]
 
 Extract `queue[]`, `runningBackground`, `_getMaxConcurrent`, `drainQueue()`, `finalizeBackgroundRun()` into a `ConcurrencyQueue` class.
+The queue stores agent IDs — not `SpawnArgs`.
+Drain calls `agent.run()` directly — no worktree setup, no args threading.
 `SettingsManager` talks to the queue directly — `notifyConcurrencyChanged()` is eliminated from `AgentManager`.
 
+- Depends on: #229
 - Target: new `src/lifecycle/concurrency-queue.ts`, `src/lifecycle/agent-manager.ts`, `src/index.ts`
 - Smell: A (tangled concerns) + C (cross-concern leak via `notifyConcurrencyChanged`)
-- Outcome: `AgentManager` loses 3 fields, 3 methods (~40 lines); scheduling is independently testable
-
-### Step 5: Push exec/registry relay deps to runner construction — [#231]
-
-`AgentManager` receives `exec` and `registry` in its constructor but only relays them to `runner.run()` via `context`.
-Move them to `ConcreteAgentRunner` construction.
-
-- Target: `src/lifecycle/agent-manager.ts`, `src/lifecycle/agent-runner.ts`, `src/index.ts`
-- Smell: C (relay-only dependencies)
-- Outcome: `AgentManager` loses 2 fields; `AgentManagerOptions` shrinks from 7 to 5 fields
+- Outcome: `AgentManager` loses 3 fields, 3 methods (~40 lines); scheduling is independently testable; queue interface is trivial (agent has everything)
 
-### Step 6: Unify resume() with RunHandle pattern — [#232]
+### Step 6: Agent.resume() with internal observer lifecycle — [#232]
 
-After #227 moves `RunHandle` ownership to the `Agent`, `resume()` on `AgentManager` becomes a 4-line delegation to `agent.resume(runner, prompt, signal)`.
-The agent manages its own observer subscription lifecycle.
+Agent has the runner from construction.
+`Agent.resume(prompt, signal)` manages its own observer subscription lifecycle using the same internal wiring as `run()`.
+`AgentManager.resume()` becomes a one-liner delegation to `agent.resume(prompt, signal)` — no manual `subscribeRecordObserver` / try-finally.
 
-- Depends on: #227, #228
-- Target: `src/lifecycle/agent-manager.ts`
+- Depends on: #229
+- Target: `src/lifecycle/agent.ts`, `src/lifecycle/agent-manager.ts`
 - Smell: A (duplicated observer subscribe/unsubscribe pattern)
-- Outcome: no manual `subscribeRecordObserver` / try-finally in the manager
+- Outcome: `AgentManager.resume()` is a 4-line delegation; observer lifecycle is Agent-internal
 
 ### Step dependency diagram
 
@@ -767,28 +825,32 @@ The agent manages its own observer subscription lifecycle.
 flowchart LR
     S1["Step 1<br/>Agent with behavior"]
     S2["Step 2<br/>async startAgent"]
-    S3["Step 3<br/>onSessionCreated observer"]
-    S4["Step 4<br/>ConcurrencyQueue"]
-    S5["Step 5<br/>relay deps"]
-    S6["Step 6<br/>resume unification"]
+    S3["Step 3<br/>runner self-contained"]
+    S4["Step 4<br/>Agent.run()"]
+    S5["Step 5<br/>ConcurrencyQueue"]
+    S6["Step 6<br/>Agent.resume()"]
 
     S1 --> S2
-    S1 --> S6
-    S2 --> S6
-    S3 ~~~ S4
-    S4 ~~~ S5
+    S2 --> S4
+    S3 --> S4
+    S4 --> S5
+    S4 --> S6
 ```
 
 ### Tracks
 
-1. **Track A — Domain model** (Steps 1, 2, 6): Agent with behavior, async runs, resume unification.
-   Sequential — each depends on the previous.
-2. **Track B — Decoupling** (Steps 3, 4, 5): independent, can proceed in parallel with Track A.
+1. **Track A — Foundation** (Step 3): Runner becomes self-contained.
+   No dependencies on other Phase 15 steps; can start immediately.
+2. **Track B — Agent lifecycle** (Steps 4, 6): Agent born complete, owns run + resume.
+   Step 4 depends on Track A + Step 2.
+   Step 6 depends on Step 4.
+3. **Track C — Scheduling** (Step 5): ConcurrencyQueue extraction.
+   Depends on Step 4 (queue drains via `agent.run()`).
 
 ## Improvement roadmap (Phase 16 — invert dependencies)
 
 Phase 16 completes the architectural inversion by removing the outbound permission bridge and the `extensions: false` / `isolated` concepts.
-It depends on Phase 15's observer pattern (#229) as the replacement mechanism.
+It depends on Phase 15's lifecycle observer (#229) as the replacement mechanism.
 
 Phase 16 is scoped but not yet broken into steps.
 Key changes:
@@ -848,7 +910,7 @@ Detailed records are preserved in per-phase history files:
 | Phase 12           | #205, #206, #207, #208                                     | renderWidgetLines, showAgentDetail, widget update, shared test fixtures                                                                                  |
 | Phase 13           | #214, #215, #216, #217, #218, #219                         | Closure-to-class, buildParentContext, startAgent decomp, overwrite guard, settings SDK, test duplication                                                 |
 | Phase 14           | #237, #238, #239, #242                                     | Remove disallowed_tools, remove extensions filtering, collapse filterActiveTools, rename Agent to subagent                                               |
-| Phase 15           | #227, #228, #229, #230, #231, #232                         | Agent domain model, async startAgent, onSessionCreated observer, ConcurrencyQueue, relay deps, resume unification                                        |
+| Phase 15           | #227, #228, #231, #229, #230, #232                         | Agent domain model, async startAgent, runner self-contained, Agent.run(), ConcurrencyQueue, Agent.resume()                                               |
 
 The remaining open issue is #22 (parent-session resolution), a cross-extension track that does not gate the structural work.
 

package/docs/plans/0228-async-start-agent-dissolve-run-handle.md

@@ -0,0 +1,288 @@
+---
+issue: 228
+issue_title: "Convert startAgent to async/await, move run lifecycle to Agent (Phase 15, Step 2)"
+---
+
+# Convert startAgent to async/await, dissolve RunHandle into Agent
+
+## Problem Statement
+
+`startAgent` is synchronous and uses `.then()`/`.catch()` to handle the runner promise.
+This forces a promise-chain callback style even though `Agent` (as of #227) already owns per-agent behavior.
+
+`RunHandle` is a private class in `agent-manager.ts` that does 6 things — 5 of which are Agent concerns (status transitions, worktree cleanup, execution state updates, listener lifecycle, signal wiring).
+The only non-Agent concern is `onFinished`, a callback that connects to the manager's concurrency queue drain.
+
+`resume()` duplicates the same pattern manually: subscribe observer, try/catch with `markCompleted`/`markError`, finally unsub.
+Issue #232 wants to unify resume with the run lifecycle, and the architecture doc says "resume becomes a 4-line delegation."
+If we just move `RunHandle` to `Agent` as a separate class, `resume()` still can't use it naturally — the signatures differ.
+But if we dissolve `RunHandle` into Agent methods, both paths use the same primitives.
+
+## Goals
+
+- Zero `.then()`/`.catch()` in `agent-manager.ts`.
+- Dissolve `RunHandle` into Agent methods: `completeRun`, `failRun`, `wireSignal`, `attachObserver`, `releaseListeners`, `onRunFinished` setter.
+- `startAgent` is a straightforward async method: setup → await → handle result.
+- `spawn()` assigns `record.promise = this.startAgent(...)`.
+- Prepare the ground for #232 (resume unification) by giving Agent the run lifecycle primitives that `resume()` can reuse.
+
+## Non-Goals
+
+- **Resume unification** — deferred to #232.
+  That issue will use the new Agent methods to simplify `AgentManager.resume()`.
+- **`onSessionCreated` observer** — deferred to #229.
+  The `onSessionCreated` callback in `startAgent` stays as-is.
+- **`ConcurrencyQueue` extraction** — deferred to #230.
+- **Relay deps** — deferred to #231.
+
+## Background
+
+### Relevant modules
+
+| Module                                 | LOC | Relationship to this change                                   |
+| -------------------------------------- | --- | ------------------------------------------------------------- |
+| `src/lifecycle/agent-manager.ts`       | 492 | Loses `RunHandle` class (~85 LOC), `startAgent` becomes async |
+| `src/lifecycle/agent.ts`               | 260 | Gains run lifecycle methods (~80 LOC)                         |
+| `src/lifecycle/agent-runner.ts`        | —   | Exports `RunResult` type, now imported by `agent.ts`          |
+| `test/lifecycle/agent.test.ts`         | 501 | Gains ~120 LOC of run lifecycle tests                         |
+| `test/lifecycle/agent-manager.test.ts` | 768 | One assertion update (`Promise<void>`)                        |
+
+### What RunHandle does today
+
+| Concern                                                                | RunHandle method     | Who should own it                               |
+| ---------------------------------------------------------------------- | -------------------- | ----------------------------------------------- |
+| Listener lifecycle (unsub + detachFn)                                  | `releaseListeners()` | Agent — per-run cleanup handles                 |
+| Run completion (worktree cleanup, status transition, execution update) | `complete(result)`   | Agent — all state mutations target Agent fields |
+| Run failure (error marking, best-effort worktree cleanup)              | `fail(err)`          | Agent — same                                    |
+| Signal wiring (parent abort → child abort)                             | `wireSignal()`       | Agent — per-run handle, released on completion  |
+| Observer attachment (session event subscription)                       | `attachObserver()`   | Agent — per-run handle, released on completion  |
+| onFinished callback (concurrency drain)                                | `fireOnFinished()`   | Manager concern, but just a stored `() => void` |
+
+Five of six are Agent concerns.
+RunHandle reaches into `this.record` for every operation and talks through `this.record.worktreeState` to a stranger.
+
+### Dependency flow (no cycles)
+
+`agent.ts` gains a type-only import of `RunResult` from `agent-runner.ts`.
+`agent-runner.ts` imports from `agent-manager.ts` (not `agent.ts`), so no cycle is created.
+
+### Constraints from AGENTS.md
+
+- `promise` type change from `Promise<string>` to `Promise<void>` is internal — `Agent` is not exported from `package.json`.
+- Worktree setup hoist preserves the synchronous-throw contract in `spawn()` (callers rely on catching `isolation: "worktree"` errors synchronously).
+
+## Design Overview
+
+### Dissolve RunHandle into Agent methods
+
+Agent gains per-run listener fields and run lifecycle methods:
+
+```typescript
+class Agent {
+  // --- Per-run listener state (released on completion or resume reset) ---
+  private _unsub?: () => void;
+  private _detachFn?: () => void;
+  private _onRunFinished?: () => void;
+
+  /** Wire a parent AbortSignal so it stops this agent when fired. */
+  wireSignal(signal: AbortSignal | undefined, onAbort: () => void): void;
+
+  /** Store the record-observer unsubscribe handle. */
+  attachObserver(unsub: () => void): void;
+
+  /** Release observer + signal listener handles. */
+  releaseListeners(): void;
+
+  /** Set the callback fired once when the run finishes (for concurrency drain). */
+  setOnRunFinished(fn: () => void): void;
+
+  /** Complete a run: release listeners, worktree cleanup, status transition,
+      execution update, fire onRunFinished. */
+  completeRun(result: RunResult, worktrees: WorktreeManager): void;
+
+  /** Fail a run: mark error, release listeners, best-effort worktree cleanup,
+      fire onRunFinished. */
+  failRun(err: unknown, worktrees: WorktreeManager): void;
+}
+```
+
+`completeRun` and `failRun` take `worktrees: WorktreeManager` as a parameter rather than storing it on Agent.
+Worktrees are only needed at run end — storing the reference would widen Agent's dependency surface for a single use.
+
+Consumer call-site after the change (`startAgent`):
+
+```typescript
+record.setOnRunFinished(
+  options.isBackground ? () => this.finalizeBackgroundRun(record) : undefined,
+);
+record.wireSignal(options.signal, () => this.abort(id));
+try {
+  const result = await this.runner.run(...);
+  record.completeRun(result, this.worktrees);
+} catch (err) {
+  record.failRun(err, this.worktrees);
+}
+```
+
+### Narrow `promise` to `Promise<void>`
+
+The resolved string value of `record.promise` is dead — every consumer just `await`s it and reads `record.result`.
+One test asserts `resolves.toBe("done")`; all others use `await record.promise`.
+Narrowing to `Promise<void>` first makes the async conversion clean (async `startAgent` naturally returns `Promise<void>`).
+
+### Hoist worktree setup from `startAgent` to callers
+
+`record.setupWorktree()` can throw synchronously (strict isolation failure).
+`spawn()` catches this and removes the orphan record.
+`drainQueue()` catches it and marks the record as errored.
+
+If `startAgent` becomes `async`, synchronous throws become rejected promises — neither caller catches them.
+Fix: move `record.setupWorktree()` into the callers' existing try-catch blocks before calling async `startAgent`.
+`startAgent` reads `record.worktreeState?.path` for the cwd instead.
+
+### `resetForResume` releases listeners
+
+After dissolution, `resetForResume` must call `releaseListeners()` and clear `_onRunFinished` to prevent stale handles from a previous run leaking into the resumed run.
+
+## Module-Level Changes
+
+### `src/lifecycle/agent.ts`
+
+1. Add per-run listener fields: `_unsub`, `_detachFn`, `_onRunFinished`.
+2. Add `wireSignal(signal, onAbort)` — logic from `RunHandle.wireSignal`.
+3. Add `attachObserver(unsub)` — logic from `RunHandle.attachObserver`.
+4. Add `releaseListeners()` — logic from `RunHandle.releaseListeners` (public).
+5. Add `setOnRunFinished(fn)` — stores the callback.
+6. Add private `fireOnRunFinished()` — idempotent clear-then-call pattern from `RunHandle.fireOnFinished`.
+7. Add `completeRun(result, worktrees)` — logic from `RunHandle.complete`, returns `void` (not `string`).
+8. Add `failRun(err, worktrees)` — logic from `RunHandle.fail`.
+9. Update `resetForResume` — call `releaseListeners()` and clear `_onRunFinished`.
+10. Change `promise` type from `Promise<string>` to `Promise<void>` (on both `AgentInit` and the class field).
+11. Add imports: `type RunResult` from `agent-runner`, `debugLog` from `debug`.
+
+### `src/lifecycle/agent-manager.ts`
+
+1. Delete `RunHandle` class (~85 lines).
+2. Remove `import type { RunResult }` (moved to `agent.ts`; `AgentRunner` import stays).
+3. Convert `startAgent` to `async`, returning `Promise<void>`.
+4. Replace RunHandle creation with Agent method calls: `record.setOnRunFinished(...)`, `record.wireSignal(...)`.
+5. Replace `handle.attachObserver(...)` with `record.attachObserver(...)` in `onSessionCreated`.
+6. Replace `.then()`/`.catch()` chain with `try { await ...; record.completeRun(...) } catch { record.failRun(...) }`.
+7. Remove `record.promise = this.runner.run(...)` assignment — `record.promise` is now assigned by `spawn`/`drainQueue`.
+8. In `spawn()`: hoist `record.setupWorktree(...)` before `startAgent` call (inside existing try-catch); assign `record.promise = this.startAgent(...)`.
+9. In `drainQueue()`: hoist `record.setupWorktree(...)` before `startAgent` call (inside existing try-catch); assign `record.promise = this.startAgent(...)`.
+10. In `startAgent`: remove `record.setupWorktree()` call; read `record.worktreeState?.path` for cwd.
+11. Update `waitForAll` filter: `Promise<string>` → `Promise<void>`.
+
+### `test/lifecycle/agent.test.ts`
+
+1. Add `describe("Agent — completeRun")` — status transitions (completed/aborted/steered), worktree cleanup with branch append, execution state update, `onRunFinished` fires once, listeners released.
+2. Add `describe("Agent — failRun")` — marks error, best-effort worktree cleanup, `onRunFinished` fires once, listeners released.
+3. Add `describe("Agent — wireSignal")` — connects parent signal to abort callback, `releaseListeners` detaches.
+4. Add `describe("Agent — attachObserver / releaseListeners")` — stores unsub, calls it on release, idempotent.
+5. Update `describe("Agent — resetForResume")` — verify listeners are released and `_onRunFinished` is cleared.
+
+### `test/lifecycle/agent-manager.test.ts`
+
+1. Update one assertion: `resolves.toBe("done")` → `resolves.toBeUndefined()`.
+
+### `packages/pi-subagents/docs/architecture/architecture.md`
+
+1. Update Phase 15 smell table — mark `startAgent` callback row as resolved.
+2. Update Step 2 description to note RunHandle dissolution (not just async conversion).
+3. Update Step 6 (#232) description — RunHandle no longer exists; Agent already has `completeRun`/`failRun`/`releaseListeners` that `resume()` can use directly.
+
+## Test Impact Analysis
+
+### New unit tests enabled by the dissolution
+
+1. **`Agent.completeRun()`** — isolated tests for run completion logic (status transitions based on `RunResult` flags, worktree cleanup, execution update, onRunFinished firing) without needing a full `AgentManager` scaffold with a mock runner.
+2. **`Agent.failRun()`** — isolated tests for error handling and best-effort cleanup.
+3. **`Agent.wireSignal()` / `Agent.attachObserver()` / `Agent.releaseListeners()`** — isolated tests for listener lifecycle without spawning a real agent.
+
+These behaviors were previously only testable through `AgentManager` integration tests that required setting up a mock runner, worktrees, and observer.
+
+### Existing tests that must stay
+
+1. All `AgentManager — spawn/spawnAndWait` tests — they verify the full spawn flow including async orchestration.
+2. All worktree isolation tests — they verify the synchronous-throw contract in `spawn()`.
+3. All queue/concurrency tests — they verify the manager's orchestration around `drainQueue`.
+4. All completion/notification tests — they verify end-to-end flow through the observer.
+
+### Existing tests that change
+
+1. One assertion in `agent-manager.test.ts`: `resolves.toBe("done")` → `resolves.toBeUndefined()` (promise type narrowing).
+
+## TDD Order
+
+1. **Narrow `Agent.promise` from `Promise<string>` to `Promise<void>`**
+   - Change `AgentInit.promise` and `Agent.promise` field types.
+   - In `startAgent`: wrap `.then()` callback body in braces (discard `handle.complete` return); remove `return ""` from `.catch()` callback.
+   - Update `waitForAll` filter type guard.
+   - Update one test assertion: `resolves.toBe("done")` → `resolves.toBeUndefined()`.
+   - Run `pnpm run check` + `pnpm vitest run`.
+   - Commit: `refactor(pi-subagents): narrow Agent.promise to Promise<void>`
+
+2. **Red/Green: add run lifecycle methods to Agent**
+   - Red: add tests in `agent.test.ts` for `completeRun`, `failRun`, `wireSignal`, `attachObserver`/`releaseListeners`, `resetForResume` listener cleanup.
+   - Green: implement the methods on `Agent` — `wireSignal`, `attachObserver`, `releaseListeners`, `setOnRunFinished`, `fireOnRunFinished`, `completeRun`, `failRun`; update `resetForResume`.
+   - Add `import type { RunResult }` and `import { debugLog }` to `agent.ts`.
+   - Run `pnpm run check` + `pnpm vitest run`.
+   - Commit: `feat(pi-subagents): add run lifecycle methods to Agent`
+
+3. **Replace RunHandle with Agent methods in `startAgent`, delete RunHandle**
+   - Replace `new RunHandle(record, this.worktrees, onFinished)` with `record.setOnRunFinished(onFinished)`.
+   - Replace `handle.wireSignal(...)` with `record.wireSignal(...)`.
+   - Replace `handle.attachObserver(...)` with `record.attachObserver(...)`.
+   - Replace `handle.complete(result)` with `record.completeRun(result, this.worktrees)`.
+   - Replace `handle.fail(err)` with `record.failRun(err, this.worktrees)`.
+   - Delete `RunHandle` class.
+   - Remove `import type { RunResult }` from `agent-manager.ts` (moved to `agent.ts`).
+   - Run `pnpm run check` + `pnpm vitest run`.
+   - Commit: `refactor(pi-subagents): replace RunHandle with Agent run lifecycle methods`
+
+4. **Hoist worktree setup from `startAgent` to callers**
+   - In `spawn()`: move `record.setupWorktree(this.worktrees, options.isolation)` before `this.startAgent()`, inside the existing try-catch.
+   - In `drainQueue()`: move `record.setupWorktree(this.worktrees, next.args.options.isolation)` before `this.startAgent()`, inside its try-catch.
+   - In `startAgent`: remove `record.setupWorktree()` call; use `record.worktreeState?.path` for `context.cwd`.
+   - Existing worktree isolation tests pass unchanged.
+   - Run `pnpm run check` + `pnpm vitest run`.
+   - Commit: `refactor(pi-subagents): hoist worktree setup from startAgent to callers`
+
+5. **Convert `startAgent` to async/await**
+   - Make `startAgent` async, returning `Promise<void>`.
+   - Replace `.then()`/`.catch()` chain with `try { const result = await this.runner.run(...); record.completeRun(result, this.worktrees); } catch (err) { record.failRun(err, this.worktrees); }`.
+   - Remove `record.promise = this.runner.run(...)` assignment from inside `startAgent`.
+   - In `spawn()`: assign `record.promise = this.startAgent(id, record, args)`.
+   - In `drainQueue()`: assign `record.promise = this.startAgent(next.id, record, next.args)`.
+   - Run `pnpm run check` + `pnpm vitest run`.
+   - Commit: `refactor(pi-subagents): convert startAgent to async/await`
+
+6. **Update architecture docs**
+   - Mark Phase 15 Step 2 smell row as resolved.
+   - Update Step 2 description to note RunHandle dissolution.
+   - Update Step 6 (#232) description: RunHandle no longer exists; Agent has `completeRun`/`failRun`/`releaseListeners` that `resume()` can use directly.
+   - Commit: `docs(pi-subagents): update architecture for async startAgent`
+
+## Risks and Mitigations
+
+1. **`resetForResume` must release listeners** — If not updated, resumed agents retain stale listener handles from the previous run.
+   Mitigated by step 2 explicitly updating `resetForResume` to call `releaseListeners()` and clear `_onRunFinished`, with a test.
+
+2. **Worktree hoist changes observer-throw semantics** — Currently, if `observer.onAgentStarted()` throws inside `startAgent`, `spawn()`'s try-catch catches it and removes the record.
+   After async conversion, that throw becomes a rejected promise.
+   This is a pre-existing inconsistency (`onAgentCompleted` is already wrapped in try-catch, `onAgentStarted` is not) and observers should not throw.
+   Mitigated by noting the inconsistency; a future step could add try-catch around `onAgentStarted`.
+
+3. **Agent grows by ~80 LOC** — Dissolving RunHandle adds methods to an already-substantial class.
+   Mitigated by the fact that these methods replace logic that already operated on Agent's fields — they belong here by SRP.
+   The net effect on `agent-manager.ts` is -85 LOC (RunHandle deletion), so the total codebase shrinks.
+
+4. **`completeRun` takes `worktrees` parameter instead of storing it** — This means every caller must pass worktrees.
+   Mitigated by there being exactly two callers today (startAgent and the future resume), both of which already have access to worktrees.
+   Storing it would widen Agent's dependency surface for a single use.
+
+## Open Questions
+
+None — the design direction (dissolve rather than move) is settled.
+The `worktrees` parameter vs. stored-reference question is resolved in favor of the parameter (ISP).