refactor delegation system, add PR summary comments, and improve code quality (#334)

* refactor delegation system and add PR summary comments Delegation system: - replace mode-based delegation with select_mode → delegate two-step flow - orchestrator crafts self-contained subagent prompts (clean context — no system/repo/event instructions leak) - add role-based tool filtering via FastMCP authenticate hook (?role=subagent hides orchestrator-only tools) - add select_mode tool for orchestrator guidance per mode - add ask_question tool for lightweight research subagents - extract shared subagent lifecycle into subagent.ts (create, complete, stdout, instructions) - route set_output to per-subagent state when activeSubagentId is set - track per-subagent state (SubagentState Map) replacing boolean delegationActive flag - capture and aggregate AgentUsage across all agents (claude, codex, gemini, opencode) - write usage summary table to GitHub job summary - block built-in subagent spawning (Task for Claude, Task(*) for Cursor) - increase activity timeout from 60s to 300s (subagent thinking phases) - fix gh CLI misguidance in system prompt — explicitly forbid usage PR summary comments: - add prSummaryComment trigger (DB schema + migrations + Zod + UI toggle) - dispatch mini-effort summary job alongside PR review on pr.created - add update_pull_request_body MCP tool - add defaultEffort option to webhook dispatch Hardening: - rewrite delegate/selectMode tests with simulated state management - add toolFiltering.test.ts for role extraction, canAccess, set_output routing - remove non-null assertions for PULLFROG_TEMP_DIR (proper error throws) - use fetchWithRetry for direct tarball downloads - DRY fix for rate limit check in test runner Co-authored-by: Cursor <cursoragent@cursor.com> * fix: add type keyword to Effort import in handleWebhook.ts Co-authored-by: Cursor <cursoragent@cursor.com> * clean up delegation system, improve code quality across the codebase - simplify delegate tool to instructions + effort params with subagent lifecycle in subagent.ts - add select_mode and ask_question orchestrator-only tools with canAccess filtering - replace delegate.test.ts/selectMode.test.ts with toolFiltering.test.ts (live MCP integration) - add set_output routing for subagent context and AgentUsage tracking across all agents - add PR summary comment trigger (schema, UI, webhook dispatch with silent flag) - add update_pull_request_body MCP tool - fix changed-agents.sh to always include claude canary for non-agent action changes - fix cursor pagination bug in getSelectedInstallationReposPage - remove destructuring patterns, inline type definitions, and unsafe type casts - replace non-null assertions with explicit checks in install.ts - convert multi-param functions to single param objects (postCleanup, runActionLocal, etc.) - use isHttpError helper in API routes instead of catch-any patterns - add adhoc test fixtures for delegation scenarios (context isolation, error handling, synthesis, etc.) Co-authored-by: Cursor <cursoragent@cursor.com> * no subagent mutation, one mcp per subagent * address review feedback: parallel-safe usage tracking, subagent isolation, minor improvements * fix subagent state isolation: replace Object.freeze with shallow copy Object.freeze throws TypeErrors when subagent tools (checkout_pr, report_progress) write scalar properties to toolState. A shallow copy achieves the same isolation for scalar fields while allowing tools to work normally. Shared references (subagents Map, usageEntries array) remain shared for coordination. --------- Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: pullfrog[bot] <226033991+pullfrog[bot]@users.noreply.github.com>
2026-02-22 14:12:43 +00:00
parent a90743e9fe
commit cfd38d82fc
43 changed files with 3293 additions and 2158 deletions
@@ -0,0 +1,167 @@
+import { createServer } from "node:net";
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
+import { type } from "arktype";
+import { FastMCP } from "fastmcp";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import { ORCHESTRATOR_ONLY_TOOLS } from "./server.ts";
+import { execute, tool } from "./shared.ts";
+import { buildSubagentInstructions } from "./subagent.ts";
+
+// ─── unit tests for pure exported functions ─────────────────────────────
+
+describe("ORCHESTRATOR_ONLY_TOOLS", () => {
+  it("includes delegation tools", () => {
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("select_mode");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("delegate");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("ask_question");
+  });
+
+  it("includes remote-mutating tools", () => {
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("push_branch");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("push_tags");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("delete_branch");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("create_pull_request");
+    expect(ORCHESTRATOR_ONLY_TOOLS).toContain("update_pull_request_body");
+  });
+});
+
+describe("buildSubagentInstructions", () => {
+  it("returns clean-room instructions with only the orchestrator prompt", () => {
+    const prompt = "Read file.ts and fix the type error.";
+    const instructions = buildSubagentInstructions(prompt);
+    expect(instructions).toEqual({
+      full: prompt,
+      system: "",
+      user: prompt,
+      eventInstructions: "",
+      repo: "",
+      event: "",
+      runtime: "",
+    });
+  });
+});
+
+// ─── per-server tool isolation integration test ─────────────────────────
+// demonstrates the architecture: orchestrator and subagent get separate servers
+
+function getRandomPort(): Promise<number> {
+  return new Promise((resolve, reject) => {
+    const srv = createServer();
+    srv.listen(0, "127.0.0.1", () => {
+      const addr = srv.address();
+      if (!addr || typeof addr === "string") return reject(new Error("bad address"));
+      const port = addr.port;
+      srv.close(() => resolve(port));
+    });
+  });
+}
+
+async function connectMcpClient(url: string): Promise<Client> {
+  const transport = new StreamableHTTPClientTransport(new URL(url));
+  const client = new Client({ name: "test-client", version: "0.0.1" });
+  // @ts-expect-error — exactOptionalPropertyTypes mismatch: SDK Transport.sessionId?: string vs StreamableHTTPClientTransport getter returning string | undefined
+  await client.connect(transport);
+  return client;
+}
+
+function mockTool(name: string, description: string) {
+  return tool({
+    name,
+    description,
+    parameters: type({ value: "string" }),
+    execute: execute(async () => ({ ok: true })),
+  });
+}
+
+describe("per-server tool isolation - integration", () => {
+  let orchestratorServer: FastMCP;
+  let subagentServer: FastMCP;
+  let orchestratorUrl: string;
+  let subagentUrl: string;
+  const clients: Client[] = [];
+
+  beforeAll(async () => {
+    const [orchestratorPort, subagentPort] = await Promise.all([getRandomPort(), getRandomPort()]);
+    orchestratorUrl = `http://127.0.0.1:${orchestratorPort}/mcp`;
+    subagentUrl = `http://127.0.0.1:${subagentPort}/mcp`;
+
+    // orchestrator gets ALL tools (common + delegation + remote mutation)
+    orchestratorServer = new FastMCP({ name: "orchestrator", version: "0.0.1" });
+    orchestratorServer.addTool(mockTool("file_read", "read a file"));
+    orchestratorServer.addTool(mockTool("git", "run git commands"));
+    orchestratorServer.addTool(mockTool("set_output", "set output"));
+    orchestratorServer.addTool(mockTool("select_mode", "select a mode"));
+    orchestratorServer.addTool(mockTool("delegate", "delegate a task"));
+    orchestratorServer.addTool(mockTool("ask_question", "ask a question"));
+    orchestratorServer.addTool(mockTool("push_branch", "push branch"));
+    orchestratorServer.addTool(mockTool("create_pull_request", "create PR"));
+
+    // subagent gets ONLY common tools (no delegation, no remote mutation)
+    subagentServer = new FastMCP({ name: "subagent", version: "0.0.1" });
+    subagentServer.addTool(mockTool("file_read", "read a file"));
+    subagentServer.addTool(mockTool("git", "run git commands"));
+    subagentServer.addTool(mockTool("set_output", "set output"));
+
+    await Promise.all([
+      orchestratorServer.start({
+        transportType: "httpStream",
+        httpStream: { port: orchestratorPort, host: "127.0.0.1", endpoint: "/mcp" },
+      }),
+      subagentServer.start({
+        transportType: "httpStream",
+        httpStream: { port: subagentPort, host: "127.0.0.1", endpoint: "/mcp" },
+      }),
+    ]);
+  });
+
+  afterAll(async () => {
+    for (const client of clients) {
+      try {
+        await client.close();
+      } catch {
+        // best-effort cleanup
+      }
+    }
+    await Promise.all([orchestratorServer.stop(), subagentServer.stop()]);
+  });
+
+  it("orchestrator sees all tools including delegation and mutation", async () => {
+    const client = await connectMcpClient(orchestratorUrl);
+    clients.push(client);
+    const result = await client.listTools();
+    const names = result.tools.map((t) => t.name);
+    expect(names).toContain("select_mode");
+    expect(names).toContain("delegate");
+    expect(names).toContain("ask_question");
+    expect(names).toContain("push_branch");
+    expect(names).toContain("create_pull_request");
+    expect(names).toContain("file_read");
+    expect(names).toContain("git");
+    expect(names).toContain("set_output");
+    expect(names.length).toBe(8);
+  });
+
+  it("subagent cannot see delegation or mutation tools", async () => {
+    const client = await connectMcpClient(subagentUrl);
+    clients.push(client);
+    const result = await client.listTools();
+    const names = result.tools.map((t) => t.name);
+    expect(names).not.toContain("select_mode");
+    expect(names).not.toContain("delegate");
+    expect(names).not.toContain("ask_question");
+    expect(names).not.toContain("push_branch");
+    expect(names).not.toContain("create_pull_request");
+  });
+
+  it("subagent sees only common tools", async () => {
+    const client = await connectMcpClient(subagentUrl);
+    clients.push(client);
+    const result = await client.listTools();
+    const names = result.tools.map((t) => t.name);
+    expect(names).toContain("file_read");
+    expect(names).toContain("git");
+    expect(names).toContain("set_output");
+    expect(names.length).toBe(3);
+  });
+});