fix(sessions): clarify delegation tool guidance

.changeset/clarify-session-delegation.md

@@ -0,0 +1,5 @@
+---
+"@jmfederico/pi-web": patch
+---
+
+Clarify agent instructions so independent sessions are created only when explicitly requested and tracked subsessions remain part of the current task.

src/server/sessions/spawnSessionTool.test.ts

@@ -20,11 +20,12 @@ describe("createSpawnSessionToolDefinition", () => {
     expect(result.content[0]).toMatchObject({ type: "text", text: "Started independent session new-1 in /repos/a-feature." });
   });
 
-  it("describes the independent-session capability without workflow policy", () => {
+  it("describes a fully independent session and restricts it to explicit requests", () => {
     const tool = createSpawnSessionToolDefinition("/repos/a", { spawn: vi.fn() });
 
-    expect(tool.description).toBe("Start a new independent pi-web session and send it an initial prompt. The session is not tracked by the caller, can be opened by a human, and runs without returning its later output to the caller.");
-    expect(tool.description).not.toMatch(/use this|continue work|follow a plan|relay/i);
+    expect(tool.description).toBe("Start a fully independent session; its transcript and results are unavailable here. Use only when the user or active workflow explicitly requests a separate session.");
+    expect(tool.promptSnippet).toBe("spawn_session: independent session; results unavailable here; explicit requests only");
+    expect(tool.description).not.toMatch(/subsession|child|parent/i);
   });
 
   it("forwards omitted cwd as undefined and omits a missing dispatching model", async () => {

src/server/sessions/spawnSessionTool.ts

@@ -41,8 +41,8 @@ export function createSpawnSessionToolDefinition(spawningCwd: string, deps: Spaw
   return defineTool<typeof SpawnSessionParams, SpawnSessionToolDetails>({
     name: "spawn_session",
     label: "Spawn session",
-    description: "Start a new independent pi-web session and send it an initial prompt. The session is not tracked by the caller, can be opened by a human, and runs without returning its later output to the caller.",
-    promptSnippet: "spawn_session: start a new independent session with a first prompt",
+    description: "Start a fully independent session; its transcript and results are unavailable here. Use only when the user or active workflow explicitly requests a separate session.",
+    promptSnippet: "spawn_session: independent session; results unavailable here; explicit requests only",
     parameters: SpawnSessionParams,
     async execute(_toolCallId, params, _signal, _onUpdate, ctx) {
       // Failures throw: the agent loop turns the thrown message into an error

src/server/sessions/spawnSubsessionTool.test.ts

@@ -35,7 +35,7 @@ function tools(deps: Partial<SubsessionToolDeps>) {
 }
 
 function workingGuidance(sessionId: string): string {
-  return `Subsession ${sessionId} is working; partial output is withheld. Continue independent work, or call yield_to_subsessions alone and last at the join point. Completion notices wake you; do not poll.`;
+  return `Subsession ${sessionId} is working; partial output is withheld. Continue other work, or call yield_to_subsessions alone and last at the join point. Completion notices wake you; do not poll.`;
 }
 
 function firstText(content: readonly (TextContent | ImageContent)[]): string {
@@ -62,16 +62,17 @@ describe("createSubsessionToolDefinitions", () => {
     expect(firstText(result.content)).toContain("Started tracked subsession child-1");
   });
 
-  it("guides the parent to continue independent work and use the explicit join action", async () => {
+  it("describes tracked child work whose result remains available to the parent", async () => {
     const { spawn: spawnTool } = tools({
       spawn: vi.fn(() => Promise.resolve({ sessionId: "child-1", cwd: "/repos/a-feature" })),
     });
 
-    expect(spawnTool.description).toBe("Start a tracked child and return immediately. Continue independent work, then use yield_to_subsessions at the join point. Completion notices wake you; do not poll.");
-    expect(spawnTool.promptSnippet).toBe("spawn_subsession: tracked parallel work; continue, then join with yield_to_subsessions");
+    expect(spawnTool.description).toBe("Start a tracked child session to carry out part of the current task and return immediately. Its transcript and result are available here after it finishes.");
+    expect(spawnTool.promptSnippet).toBe("spawn_subsession: tracked child work for the current task; result available after completion");
+    expect(spawnTool.description).not.toMatch(/spawn_session|fully independent/i);
 
     const result = await spawnTool.execute("call-contract", { prompt: "do it" }, undefined, undefined, ctxFor("parent-1", undefined));
-    expect(firstText(result.content)).toBe("Started tracked subsession child-1 in /repos/a-feature. Continue independent work, then join with yield_to_subsessions; do not poll.");
+    expect(firstText(result.content)).toBe("Started tracked subsession child-1 in /repos/a-feature. Continue other work, then join with yield_to_subsessions; do not poll.");
   });
 
   it("distinguishes status inspection from yielding in tool metadata", () => {
@@ -92,9 +93,7 @@ describe("createSubsessionToolDefinitions", () => {
     expect(yieldTool.description).toBe("At a join point, end this run while tracked children work; completion notices wake you. If none work, continue. Call alone and last; do not poll.");
     expect(yieldTool.promptSnippet).toBe("yield_to_subsessions: end the run at a join point; call alone and last");
     expect(yieldTool.promptGuidelines).toEqual([
-      "After independent work, yield only at a join point; use spawn_session for fire-and-forget work.",
-      "Call alone and last; a mixed tool batch may continue the run.",
-      "Completion notices wake you; do not poll inspection tools.",
+      "After calling spawn_subsession, you can continue with other work. At the join point, call yield_to_subsessions alone and last; completion notices wake you, so do not poll.",
     ]);
   });
 

src/server/sessions/spawnSubsessionTool.ts

@@ -120,7 +120,7 @@ function statusLine(summary: SubsessionSummary): string {
 }
 
 function workingInspectionGuidance(sessionId: string): string {
-  return `Subsession ${sessionId} is working; partial output is withheld. Continue independent work, or call yield_to_subsessions alone and last at the join point. Completion notices wake you; do not poll.`;
+  return `Subsession ${sessionId} is working; partial output is withheld. Continue other work, or call yield_to_subsessions alone and last at the join point. Completion notices wake you; do not poll.`;
 }
 
 function renderEntry(entry: TranscriptEntry): string {
@@ -174,18 +174,18 @@ function renderTranscript(result: SubsessionReadResult): string {
  * Tools that let an agent spawn *tracked* child sessions, inspect them, and
  * explicitly yield at a join point.
  *
- * Unlike `spawn_session` (fire-and-forget peers), a subsession records its
- * parent in its session header, the parent is notified when it stops working,
- * and the parent may read its transcript/result. The tools are constructed
- * per-session, carrying the spawning cwd for project-scope validation; the
- * parent's identity is taken from the live extension context at call time.
+ * A subsession records its parent in its session header, the parent is notified
+ * when it stops working, and the parent may read its transcript/result. The
+ * tools are constructed per-session, carrying the spawning cwd for
+ * project-scope validation; the parent's identity is taken from the live
+ * extension context at call time.
  */
 export function createSubsessionToolDefinitions(spawningCwd: string, deps: SubsessionToolDeps) {
   const spawnTool = defineTool<typeof SpawnSubsessionParams, SpawnSubsessionResult>({
     name: "spawn_subsession",
     label: "Spawn subsession",
-    description: "Start a tracked child and return immediately. Continue independent work, then use yield_to_subsessions at the join point. Completion notices wake you; do not poll.",
-    promptSnippet: "spawn_subsession: tracked parallel work; continue, then join with yield_to_subsessions",
+    description: "Start a tracked child session to carry out part of the current task and return immediately. Its transcript and result are available here after it finishes.",
+    promptSnippet: "spawn_subsession: tracked child work for the current task; result available after completion",
     parameters: SpawnSubsessionParams,
     async execute(_toolCallId, params, _signal, _onUpdate, ctx) {
       const parentSessionId = ctx.sessionManager.getSessionId();
@@ -199,7 +199,7 @@ export function createSubsessionToolDefinitions(spawningCwd: string, deps: Subse
         ...(ctx.model === undefined ? {} : { model: ctx.model }),
       });
       return {
-        content: [{ type: "text", text: `Started tracked subsession ${result.sessionId} in ${result.cwd}. Continue independent work, then join with yield_to_subsessions; do not poll.` }],
+        content: [{ type: "text", text: `Started tracked subsession ${result.sessionId} in ${result.cwd}. Continue other work, then join with yield_to_subsessions; do not poll.` }],
         details: result,
       };
     },
@@ -270,9 +270,7 @@ export function createSubsessionToolDefinitions(spawningCwd: string, deps: Subse
     description: "At a join point, end this run while tracked children work; completion notices wake you. If none work, continue. Call alone and last; do not poll.",
     promptSnippet: "yield_to_subsessions: end the run at a join point; call alone and last",
     promptGuidelines: [
-      "After independent work, yield only at a join point; use spawn_session for fire-and-forget work.",
-      "Call alone and last; a mixed tool batch may continue the run.",
-      "Completion notices wake you; do not poll inspection tools.",
+      "After calling spawn_subsession, you can continue with other work. At the join point, call yield_to_subsessions alone and last; completion notices wake you, so do not poll.",
     ],
     parameters: YieldToSubsessionsParams,
     async execute(_toolCallId, _params, _signal, _onUpdate, ctx) {