pelikhan · Oct 8, 2026 · Oct 8, 2026 · Oct 8, 2026 · Oct 8, 2026 · github-actions
diff --git a/.changeset/fix-pi-smoke-completion.md b/.changeset/fix-pi-smoke-completion.md
diff --git a/.github/workflows/smoke-pi-sub-agents.lock.yml b/.github/workflows/smoke-pi-sub-agents.lock.yml
diff --git a/.github/workflows/smoke-pi-sub-agents.md b/.github/workflows/smoke-pi-sub-agents.md
@@ -40,8 +40,8 @@ This is the Pi variant of `smoke-copilot-sub-agents.md`. Keep output short.
 1. Use the managed `subagent` tool exactly once per declared agent, sequentially:
    - `{"agent":"haiku-whoami","task":"who am i?"}`
    - `{"agent":"mini-whoami","task":"who am i?"}`
-   - `{"agent":"nano-whoami","task":"who am i?"}`
-2. Check the exact responses: `claude-haiku-4.5`, `gpt-5-mini`, and `gpt-5-nano`.
+   - `{"agent":"compact-whoami","task":"who am i?"}`
+2. Check the exact responses: `claude-haiku-4.5`, `gpt-5-mini`, and `gpt-4o-mini`.
 3. Mark a missing tool, failed delegation, or unexpected response as FAIL. Do not
    answer on behalf of a sub-agent or substitute a different agent.
 4. Do not make other tool calls except the staged `create_issue` report below.
@@ -55,36 +55,40 @@ Create a staged issue titled **"Smoke Test: Pi Sub Agents - ${{ github.run_id }}
 
 These fixed responses test delegation, not a model's ability to identify itself.
 The dispatch records and inference model usage provide model-selection evidence.
+Provider-qualified model IDs pin the exact models rather than resolving aliases.
 
 ## agent: `haiku-whoami`
 ---
 description: Returns the Haiku model identity for smoke testing
-model: claude-haiku-4.5
+model: copilot/claude-haiku-4.5
 ---
-When asked `who am i?`, reply with exactly:
+When asked `who am i?`, reply with this single plain-text line, without Markdown
+or backticks:
 
-`claude-haiku-4.5`
+claude-haiku-4.5
 
-No extra words, punctuation, or formatting.
+Do not call tools. No extra words, punctuation, or formatting.
 
 ## agent: `mini-whoami`
 ---
 description: Returns the GPT-5 mini model identity for smoke testing
-model: gpt-5-mini
+model: copilot/gpt-5-mini
 ---
-When asked `who am i?`, reply with exactly:
+When asked `who am i?`, reply with this single plain-text line, without Markdown
+or backticks:
 
-`gpt-5-mini`
+gpt-5-mini
 
-No extra words, punctuation, or formatting.
+Do not call tools. No extra words, punctuation, or formatting.
 
-## agent: `nano-whoami`
+## agent: `compact-whoami`
 ---
-description: Returns the GPT-5 nano model identity for smoke testing
-model: gpt-5-nano
+description: Returns the GPT-4o mini model identity for smoke testing
+model: copilot/gpt-4o-mini
 ---
-When asked `who am i?`, reply with exactly:
+When asked `who am i?`, reply with this single plain-text line, without Markdown
+or backticks:
 
-`gpt-5-nano`
+gpt-4o-mini
 
-No extra words, punctuation, or formatting.
+Do not call tools. No extra words, punctuation, or formatting.
diff --git a/actions/setup/js/pi_steering_extension.cjs b/actions/setup/js/pi_steering_extension.cjs
@@ -60,8 +60,8 @@ function piSteeringExtension(pi) {
     process.stderr.write(`[gh-aw/steering] Session started. timeout=${config.timeoutMinutes}min, warn<${config.timeWarningMinutes}min, critical<${config.timeCriticalMinutes}min\n`);
   });
 
-  pi.on("turn_end", async (/** @type {any} */ _event, /** @type {any} */ ctx) => {
-    if (startTime === undefined) {
+  pi.on("turn_end", async (/** @type {any} */ event, /** @type {any} */ ctx) => {
+    if (startTime === undefined || event.message?.stopReason !== "toolUse") {
       return;
     }
 

diff --git a/actions/setup/js/pi_steering_extension.test.cjs b/actions/setup/js/pi_steering_extension.test.cjs
@@ -49,13 +49,13 @@ describe("Pi timeout steering", () => {
     const { handlers, sendUserMessage } = harness();
     await handlers.agent_start();
     vi.advanceTimersByTime(6 * 60_000);
-    await handlers.turn_end({}, {});
-    await handlers.turn_end({}, {});
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
     expect(sendUserMessage).toHaveBeenCalledTimes(1);
     expect(sendUserMessage).toHaveBeenNthCalledWith(1, expect.stringContaining("wrap up"), { deliverAs: "steer" });
     vi.advanceTimersByTime(3 * 60_000);
-    await handlers.turn_end({}, {});
-    await handlers.turn_end({}, {});
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
     expect(sendUserMessage).toHaveBeenCalledTimes(2);
     expect(sendUserMessage).toHaveBeenNthCalledWith(2, expect.stringContaining("CRITICAL"), { deliverAs: "steer" });
   });
@@ -65,7 +65,24 @@ describe("Pi timeout steering", () => {
     await handlers.agent_start();
     vi.advanceTimersByTime(9 * 60_000);
     await handlers.agent_start();
-    await handlers.turn_end({}, {});
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
+    expect(sendUserMessage).toHaveBeenCalledOnce();
+    expect(sendUserMessage).toHaveBeenCalledWith(expect.stringContaining("CRITICAL"), { deliverAs: "steer" });
+  });
+
+  it.each(["stop", "error", "aborted", "unknown", undefined])("does not steer terminal assistant turns with stopReason %s", async stopReason => {
+    const { handlers, sendUserMessage } = harness();
+    await handlers.agent_start();
+    vi.advanceTimersByTime(9 * 60_000);
+    await handlers.turn_end({ message: { role: "assistant", stopReason } }, {});
+    expect(sendUserMessage).not.toHaveBeenCalled();
+  });
+
+  it("steers only ongoing tool turns", async () => {
+    const { handlers, sendUserMessage } = harness();
+    await handlers.agent_start();
+    vi.advanceTimersByTime(9 * 60_000);
+    await handlers.turn_end({ message: { role: "assistant", stopReason: "toolUse" } }, {});
     expect(sendUserMessage).toHaveBeenCalledOnce();
     expect(sendUserMessage).toHaveBeenCalledWith(expect.stringContaining("CRITICAL"), { deliverAs: "steer" });
   });

diff --git a/actions/setup/js/pi_subagent_extension.cjs b/actions/setup/js/pi_subagent_extension.cjs
@@ -38,7 +38,11 @@ async function runPiSubagent(agent, task, ctx, signal, launch = spawn) {
     const promptPath = path.join(dir, "system.txt");
     const systemPath = process.env.GH_AW_PI_SYSTEM_PROMPT;
     const system = systemPath ? fs.readFileSync(systemPath, "utf8") + "\n\n" : "";
-    fs.writeFileSync(promptPath, system + agent.prompt, { mode: 0o600 });
+    const delegationScope =
+      "You are a delegated sub-agent, not the workflow's main agent. Complete only the delegated task and return your answer to the parent, following the declared agent's output format exactly. " +
+      "The parent is responsible for finalizing the workflow and its required safe-output reporting. Do not emit noop safe outputs or workflow-completion reports for this child session. " +
+      "Safe-output actions required to perform the delegated task remain permitted.";
+    fs.writeFileSync(promptPath, system + agent.prompt + "\n\n" + delegationScope, { mode: 0o600 });
     process.stdout.write(JSON.stringify({ type: "gh_aw_subagent_dispatch", agent: agent.name, requested_model: agent.declaredModel || agent.modelId, resolved_model: agent.modelId }) + "\n");
     return await new Promise((resolve, reject) => {
       const child = launch(process.env.GH_AW_PI_COMMAND || "pi", subagentArgs(agent, promptPath), {

diff --git a/actions/setup/js/pi_subagent_extension.test.cjs b/actions/setup/js/pi_subagent_extension.test.cjs
@@ -93,6 +93,21 @@ describe("Pi managed delegation", () => {
     await expect(runPiSubagent(agent, "Read.", { cwd: dir }, undefined, launcher({ stopReason: "error", errorMessage: "Model unavailable" }))).rejects.toThrow("Model unavailable");
   });
 
+  it("inherits workflow controls without making the child finalize the workflow", async () => {
+    const systemPath = path.join(dir, "workflow-system.txt");
+    fs.writeFileSync(systemPath, "Workflow instructions and tool restrictions.");
+    vi.stubEnv("GH_AW_PI_SYSTEM_PROMPT", systemPath);
+    const launch = launcher();
+    await runPiSubagent(agent, "Read.", { cwd: dir }, undefined, (...args) => {
+      const prompt = fs.readFileSync(args[1].at(-1), "utf8");
+      expect(prompt).toContain("Workflow instructions and tool restrictions.\n\nRead only.");
+      expect(prompt).toContain("The parent is responsible for finalizing the workflow");
+      expect(prompt).toContain("Do not emit noop safe outputs");
+      expect(prompt).toContain("Safe-output actions required to perform the delegated task remain permitted");
+      return launch(...args);
+    });
+  });
+
   it("propagates nonzero child exits", async () => {
     await expect(runPiSubagent(agent, "Read.", { cwd: dir }, undefined, launcher({}, 1))).rejects.toThrow();
   });
@@ -171,13 +186,14 @@ process.stdin.on("end", () => {
         GH_AW_SUB_AGENT_EXT: ".md",
         GH_AW_PI_SUBAGENT_ARGS: originalEnv.GH_AW_PI_SUBAGENT_ARGS,
         GH_AW_PI_TOOL_POLICY: originalEnv.GH_AW_PI_TOOL_POLICY,
+        GH_AW_PI_MODEL_ALIASES: JSON.stringify({ "gpt-5-mini": ["copilot/gpt-5*mini*"], "gpt-5-nano": ["copilot/gpt-5*nano*"] }),
       };
       vi.stubGlobal("core", { info: vi.fn() });
       const source = fs.readFileSync(new URL("../../../.github/workflows/smoke-pi-sub-agents.md", import.meta.url), "utf8");
       const parentPrompt = writeInlineSubAgents(source, dir, dir, "pi");
       expect(parentPrompt).not.toContain("## agent:");
       stagePiArtifacts(process.env.PI_CODING_AGENT_DIR);
-      const expected = ["claude-haiku-4.5", "gpt-5-mini", "gpt-5-nano"];
+      const expected = ["claude-haiku-4.5", "gpt-5-mini", "gpt-4o-mini"];
       const sdk = {
         parseFrontmatter: content => {
           const [, header, body] = content.split("---");
@@ -197,7 +213,7 @@ process.stdin.on("end", () => {
         agentDir: process.env.PI_CODING_AGENT_DIR,
         sdk,
         provider: "github-copilot",
-        catalog: expected.map(model => `github-copilot/${model}`),
+        catalog: [...expected, "gpt-5.4-mini"].map(model => `github-copilot/${model}`),
         gateway: true,
         parentModel: "github-copilot/gpt-5.3-codex",
       });
@@ -221,7 +237,7 @@ process.stdin.on("end", () => {
       process.env.GH_AW_PI_COMMAND = executable;
       for (const agent of agents) {
         const result = await runPiSubagent(agent, "who am i?", { cwd: dir });
-        expect(result.content[0].text).toBe(agent.declaredModel);
+        expect(result.content[0].text).toBe(agent.modelId);
         expect(result.details.model).toBe(agent.modelId);
       }
     } finally {

diff --git a/actions/setup/js/pi_tool_policy.cjs b/actions/setup/js/pi_tool_policy.cjs
@@ -68,6 +68,7 @@ function piToolPolicy(pi) {
       callCount = reservePiBudget("calls", callCount, limit);
       let reason;
       if (limit !== undefined && callCount > limit) reason = `Workflow tool-call budget exhausted (${limit} calls)`;
+      else if (process.env.GH_AW_PI_SUBAGENT_CHILD === "1" && /(?:^|[^a-z0-9])(?:noop|report_incomplete|create_report_incomplete_issue)$/i.test(event.toolName || "")) reason = "Workflow-completion reporting is owned by the parent agent";
       else if (event.toolName === "bash" && !isPiBashAllowed(event.input?.command, policy.bash)) reason = "Command is outside the workflow tools.bash allowlist";
       else if (policy.edit === false && ["edit", "write"].includes(event.toolName)) reason = "File editing is disabled by the workflow";
       if (!reason) return;

diff --git a/actions/setup/js/pi_tool_policy.test.cjs b/actions/setup/js/pi_tool_policy.test.cjs
@@ -102,6 +102,21 @@ describe("Pi tool policy", () => {
     for (const toolName of ["edit", "write"]) expect(handlers.tool_call({ toolName }, {})).toMatchObject({ block: true });
   });
 
+  it("blocks child workflow-completion reporting without blocking task safe outputs", () => {
+    vi.stubEnv("GH_AW_PI_SUBAGENT_CHILD", "1");
+    const child = {};
+    policyExtension({ on: (event, handler) => (child[event] = handler) });
+    for (const toolName of ["noop", "safeoutputs__report_incomplete", "safeoutputs__create_report_incomplete_issue"]) {
+      expect(child.tool_call({ toolName }, {})).toMatchObject({ block: true, reason: expect.stringContaining("parent agent") });
+    }
+    expect(child.tool_call({ toolName: "safeoutputs__create_issue" }, {})).toBeUndefined();
+
+    vi.stubEnv("GH_AW_PI_SUBAGENT_CHILD", "");
+    const parent = {};
+    policyExtension({ on: (event, handler) => (parent[event] = handler) });
+    expect(parent.tool_call({ toolName: "safeoutputs__noop" }, {})).toBeUndefined();
+  });
+
   it("aborts inference at the configured repeated-denial threshold", () => {
     const originalExitCode = process.exitCode;
     vi.stubEnv("GH_AW_PI_TOOL_POLICY", '{"bash":false}');

diff --git a/docs/src/content/docs/engines/pi.md b/docs/src/content/docs/engines/pi.md
@@ -143,6 +143,8 @@ An omitted model inherits the parent's configured model. Bare model IDs use the
 
 Children retain the workflow's inference gateway, MCP configuration, steering, bash allowlist, and edit restrictions. Tool-call and denial budgets are shared with the parent. Children use compiler-managed execution arguments rather than inheriting `engine.args` overrides. Optional automatically discovered extensions are disabled in children; only the managed infrastructure extensions are loaded. Custom drivers do not support this delegation path.
 
+Children return task results to the parent instead of emitting workflow-completion reports or `noop` safe outputs. Task-required safe-output actions remain available. Time-pressure steering applies to ongoing tool turns, not completed answers, so it does not reopen a finished child session.
+
 Dispatch logs record both the requested alias and resolved model. `gh aw audit` reports `subagent_model_not_observed` when usage evidence exists but no agent inference matches a declared model. This means the agent may not have run or its model was not honored; model presence alone does not prove delegation. Detection requests are not treated as agent-model evidence.
 
 ## Additional providers

diff --git a/pkg/cli/token_usage_declared_subagents.go b/pkg/cli/token_usage_declared_subagents.go
@@ -5,10 +5,16 @@ import (
 	"maps"
 	"os"
 	"path"
+	"regexp"
 	"slices"
 	"strings"
 )
 
+var (
+	effectiveModelDateSuffix     = regexp.MustCompile(`-[0-9]{4}(?:-[0-9]{2}-[0-9]{2}|[0-9]{4})$`)
+	claudeHyphenatedModelVersion = regexp.MustCompile(`^(claude-.+)-([0-9]+)-([0-9]+)$`)
+)
+
 func augmentDeclaredSubagentModels(runDir string, summary *TokenUsageSummary) {
 	infoPath := findAwInfoPath(runDir)
 	if infoPath == "" {
@@ -98,9 +104,22 @@ func matchesDeclaredModel(pattern, observed, provider string) bool {
 		pattern = model
 	}
 	matched, err := path.Match(pattern, observed)
+	if err == nil && matched {
+		return true
+	}
+	normalized := normalizeEffectiveModelName(observed)
+	if normalized == observed {
+		return false
+	}
+	matched, err = path.Match(pattern, normalized)
 	return err == nil && matched
 }
 
+func normalizeEffectiveModelName(model string) string {
+	model = effectiveModelDateSuffix.ReplaceAllString(model, "")
+	return claudeHyphenatedModelVersion.ReplaceAllString(model, "$1-$2.$3")
+}
+
 func generateSubagentModelFindings(summary *TokenUsageSummary) []AuditFinding {
 	if summary == nil {
 		return nil