@kendoo.agentdesk/agentdesk 0.26.0 → 0.28.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kendoo.agentdesk/agentdesk",
-  "version": "0.26.0",
+  "version": "0.28.0",
   "description": "AI team orchestrator for Claude Code — run collaborative agent sessions from your terminal",
   "type": "module",
   "bin": {
@@ -9,7 +9,6 @@
   "files": [
     "bin/",
     "cli/",
-    "prompts/",
     "README.md",
     "CHANGELOG.md"
   ],
@@ -22,11 +21,15 @@
     "server": "node server/index.mjs",
     "build": "vite build",
     "preview": "vite preview",
-    "test": "node --test tests/server.test.mjs tests/agents.test.mjs tests/homepage.test.mjs tests/sessionUtils.test.mjs tests/tracker-url.test.mjs tests/session-preflight.test.mjs tests/access.test.mjs tests/project-ownership.test.mjs tests/random-hex.test.mjs",
+    "test": "node --test tests/server.test.mjs tests/agents.test.mjs tests/homepage.test.mjs tests/sessionUtils.test.mjs tests/tracker-url.test.mjs tests/session-preflight.test.mjs tests/access.test.mjs tests/project-ownership.test.mjs tests/random-hex.test.mjs tests/projects-registry.test.mjs tests/phase-loop.test.mjs tests/proc.test.mjs tests/crypto.test.mjs tests/dotenv.test.mjs tests/update-check.test.mjs tests/setup-helpers.test.mjs tests/project-key.test.mjs tests/tracker-project.test.mjs tests/tracker-check.test.mjs tests/config.test.mjs tests/engine-env.test.mjs tests/engine-events.test.mjs tests/engine-verdict.test.mjs tests/engine-hooks.test.mjs tests/engine-agents.test.mjs tests/session-isolation.test.mjs tests/engine-session.test.mjs tests/engine-prompts.test.mjs tests/engine-schemas.test.mjs tests/engine-claude-auth.test.mjs",
+    "test:coverage": "node --test --experimental-test-coverage --test-coverage-include='cli/**' --test-coverage-include='server/**' --test-coverage-lines=60 --test-coverage-branches=62 tests/*.test.mjs",
+    "lint": "eslint .",
+    "lint:fix": "eslint . --fix",
     "lint:changelog": "node scripts/lint-changelog.mjs",
     "prepublishOnly": "node scripts/lint-changelog.mjs"
   },
   "dependencies": {
+    "@anthropic-ai/claude-agent-sdk": "0.3.266",
     "@inquirer/prompts": "^7.10.1",
     "@radix-ui/react-avatar": "^1.1.11",
     "@radix-ui/react-dialog": "^1.1.15",
@@ -51,9 +54,15 @@
     "ws": "^8.18.0"
   },
   "devDependencies": {
+    "@eslint/js": "^9.39.5",
     "@vitejs/plugin-react": "^4.5.2",
     "autoprefixer": "^10.4.21",
     "concurrently": "^9.2.0",
+    "eslint": "^9.39.5",
+    "eslint-plugin-import": "^2.32.0",
+    "eslint-plugin-react": "^7.37.5",
+    "eslint-plugin-react-hooks": "^7.1.1",
+    "globals": "^17.12.0",
     "lucide-react": "^0.577.0",
     "postcss": "^8.5.4",
     "react": "^19.1.0",

package/cli/orchestrator.mjs

@@ -1,461 +0,0 @@
-// Orchestrator — runs a single Claude process with all agents as personas
-
-import { spawn, execSync } from "child_process";
-import { existsSync, readFileSync, writeFileSync, unlinkSync } from "fs";
-import { createInterface } from "readline";
-import { join, dirname } from "path";
-import { fileURLToPath } from "url";
-import { buildPrompt, buildSoloPrompt, buildPhasedPrompt } from "./prompt.mjs";
-import { createStreamParser } from "./stream-parser.mjs";
-import { createScratchHome } from "./session-sandbox.mjs";
-import { wrapIsolatedSpawn, probeIsolation } from "./session-isolation.mjs";
-import { buildTrackerUrl } from "./tracker-url.mjs";
-import { resolveGitHubCreds, assertPushable, PreflightError } from "./session-preflight.mjs";
-import { loadDotEnv } from "./dotenv.mjs";
-
-const __dirname = dirname(fileURLToPath(import.meta.url));
-const CLI_VERSION = JSON.parse(readFileSync(join(__dirname, "../package.json"), "utf-8")).version;
-
-// Fetch decrypted tracker credentials from server
-async function fetchTrackerCredentials(projectName, apiKey, serverUrl) {
-  if (!apiKey || !serverUrl || !projectName) return {};
-  try {
-    const res = await fetch(`${serverUrl}/api/projects/${projectName}/settings/credentials`, {
-      headers: { "x-api-key": apiKey },
-      signal: AbortSignal.timeout(5000),
-    });
-    if (res.ok) return await res.json();
-  } catch {}
-  return {};
-}
-
-function timestamp() {
-  const d = new Date();
-  return [d.getHours(), d.getMinutes(), d.getSeconds()]
-    .map(n => String(n).padStart(2, "0")).join(":");
-}
-
-// Phase-specific fallback when the user hasn't set a model override.
-// REVIEW and SUMMARY default to haiku (fast + cheap — no heavy reasoning needed).
-// INTAKE/PLAN/EXECUTION default to Claude Code's default (sonnet).
-function defaultForPhase(phase) {
-  if (phase === "REVIEW" || phase === "SUMMARY") return "haiku";
-  return null;
-}
-
-// Resolve --model args for a given phase based on project settings.
-// phaseModels is { INTAKE?, PLAN?, EXECUTION?, REVIEW?, SUMMARY? } with values "opus"|"sonnet"|"haiku"|"default".
-// Returns [] when no override (so Claude Code picks its default).
-function modelArgsForPhase(phase, phaseModels) {
-  let choice = phaseModels?.[phase];
-  if (!choice || choice === "default") choice = defaultForPhase(phase);
-  if (!choice) return [];
-  return ["--model", choice];
-}
-
-// Display label for a phase's model.
-function modelLabelForPhase(phase, phaseModels) {
-  const override = phaseModels?.[phase];
-  if (override && override !== "default") return override;
-  return defaultForPhase(phase) || "default";
-}
-
-export async function runOrchestrator({
-  taskId, taskLink, description, createTask, tracker, config,
-  project, team, teamSections, sessionUrl, cwd,
-  onEvent, apiKey, serverUrl, soloAgent, childStrategy, sessionId,
-}) {
-  const trackerCreds = await fetchTrackerCredentials(project?.name, apiKey, serverUrl);
-
-  // Resolve creds from server + project .env BEFORE building the scratch
-  // HOME, so the sandbox helper writes the right token (or we fail fast
-  // when none is available — see assertPushable below).
-  const resolvedToken = resolveGitHubCreds({ cwd, trackerCreds }).GITHUB_TOKEN;
-  const resolvedCreds = { ...trackerCreds, GITHUB_TOKEN: resolvedToken || trackerCreds.GITHUB_TOKEN };
-
-  const startTime = Date.now();
-  let totalInputTokens = 0;
-  let totalOutputTokens = 0;
-  let totalSteps = 0;
-
-  let lastPhase = null;
-
-  function emit(event) {
-    if (event.type === "phase:change") lastPhase = event.phase;
-    onEvent?.({ ...event, timestamp: timestamp() });
-  }
-
-  const fullPrompt = soloAgent
-    ? buildSoloPrompt({ agentName: soloAgent, taskId, description, tracker, config, project, sessionUrl, childStrategy, cwd })
-    : buildPrompt({ taskId, taskLink, description, createTask, tracker, config, project, teamSections, sessionUrl, cwd });
-
-  emit({
-    type: "session:start",
-    taskId, taskLink,
-    title: description || taskId,
-    project: project?.name || null,
-    sessionNumber: 1,
-    agents: soloAgent ? [soloAgent] : teamSections.names,
-    cliVersion: CLI_VERSION,
-  });
-
-  // Preflight: must run AFTER session:start (so the UI has a session to
-  // bind the error to) but BEFORE we touch the scratch sandbox or spawn
-  // the child. Surfaces missing creds / SSH-only remotes as a clean
-  // session error instead of letting the agent silently fail to push.
-  try {
-    assertPushable({ cwd, creds: resolvedCreds, projectName: project?.name });
-  } catch (err) {
-    if (err instanceof PreflightError) {
-      emit({ type: "session:error", code: err.code, message: err.message });
-      emit({ type: "session:end", duration: `${((Date.now() - startTime) / 1000).toFixed(1)}s`, steps: 0, inputTokens: 0, outputTokens: 0, status: "error" });
-      return { duration: "0s", steps: 0, inputTokens: 0, outputTokens: 0, handoff: false, error: err.message };
-    }
-    throw err;
-  }
-
-  // Build a per-session scratch HOME that isolates THIS project's identity
-  // from the user's global git/gh/ssh state and from other projects'
-  // credentials. Cleanup runs at session end regardless of outcome.
-  const sandbox = createScratchHome({
-    projectId: project?.name,
-    sessionId,
-    creds: resolvedCreds,
-    commitIdentity: {
-      name: config?.identityBadge || "AgentDesk",
-      email: resolvedCreds.JIRA_EMAIL || `agentdesk@local`,
-    },
-  });
-  const env = { ...process.env, ...loadDotEnv(cwd), ...sandbox.env };
-
-  const initialPhase = soloAgent ? "EXECUTION" : "INTAKE";
-  // Non-phased single run spans all phases in one process — use EXECUTION's model
-  // as the representative choice (matches where the heavy lifting happens).
-  const representativeModel = modelLabelForPhase("EXECUTION", config?.phaseModels);
-  emit({ type: "phase:change", phase: initialPhase, model: representativeModel });
-
-  const modelArgs = modelArgsForPhase("EXECUTION", config?.phaseModels);
-  const phaseModelsResolved = config?.phaseModels && Object.keys(config.phaseModels).length
-    ? JSON.stringify(config.phaseModels)
-    : "(none — using Claude Code defaults)";
-  console.error(`[agentdesk] phaseModels: ${phaseModelsResolved}`);
-  console.error(`[agentdesk] EXECUTION model: ${modelArgs.length ? modelArgs[1] : "default (no --model flag)"}`);
-  const wrapped = wrapIsolatedSpawn({
-    cmd: "claude",
-    args: ["-p", fullPrompt, ...modelArgs, "--allowedTools", "Bash,Read,Edit,Write,Glob,Grep", "--verbose", "--output-format", "stream-json"],
-    cwd,
-    scratchHome: sandbox.home,
-    sessionId,
-  });
-  if (wrapped.isolation.kind !== "none") {
-    console.error(`[agentdesk] session isolation: ${wrapped.isolation.kind}`);
-  } else if (wrapped.isolation.reason) {
-    console.error(`[agentdesk] hard isolation unavailable (${wrapped.isolation.reason}) — scoped-env only`);
-  }
-  const child = spawn(
-    wrapped.cmd,
-    wrapped.args,
-    { stdio: ["pipe", "pipe", "inherit"], shell: false, env, cwd }
-  );
-  child.stdin.end();
-
-  const { parseLine } = createStreamParser({
-    teamNames: soloAgent ? [soloAgent] : teamSections.names,
-    callbacks: {
-      onPhaseChange({ phase }) { emit({ type: "phase:change", phase, model: representativeModel }); },
-      onModel({ model }) { emit({ type: "session:model", model }); },
-      onAgentMessage({ agent, tag, message }) { emit({ type: "agent:message", agent, tag, message }); },
-      onToolUse({ agent, tool, description }) { totalSteps++; emit({ type: "tool:use", agent, tool, description }); },
-      onToolResult({ success, summary }) { emit({ type: "tool:result", success, summary }); },
-      onSessionUpdate({ taskId: newTaskId, title }) {
-        if (newTaskId) {
-          const update = { type: "session:update", taskId: newTaskId };
-          const taskLink = buildTrackerUrl({ tracker, config, taskId: newTaskId });
-          if (taskLink) update.taskLink = taskLink;
-          emit(update);
-        }
-        if (title) emit({ type: "session:update", title });
-      },
-      onSessionEnd({ duration, steps, inputTokens, outputTokens }) {
-        totalInputTokens = inputTokens;
-        totalOutputTokens = outputTokens;
-        totalSteps = steps;
-      },
-    },
-  });
-
-  const rl = createInterface({ input: child.stdout });
-  for await (const line of rl) {
-    parseLine(line);
-  }
-
-  const exitCode = await new Promise(resolve => child.on("close", resolve));
-
-  const duration = `${((Date.now() - startTime) / 1000).toFixed(1)}s`;
-
-  // Detect limit/crash — non-zero exit without a clean session end
-  const isHandoff = exitCode !== 0;
-
-  if (isHandoff) {
-    // Write mechanical resume snapshot
-    let branch = "";
-    let diffStat = "";
-    try { branch = execSync("git branch --show-current", { cwd, encoding: "utf-8" }).trim(); } catch {}
-    try { diffStat = execSync("git diff --stat HEAD", { cwd, encoding: "utf-8" }).trim(); } catch {}
-
-    const resumePath = join(cwd, ".agentdesk-resume.md");
-    const resumeContent = [
-      `# AgentDesk Resume — ${taskId}`,
-      ``,
-      `Session: ${sessionUrl}`,
-      `Date: ${new Date().toISOString()}`,
-      `Phase: ${lastPhase || "UNKNOWN"}`,
-      `Duration: ${duration}`,
-      `Steps: ${totalSteps}`,
-      `Exit code: ${exitCode}`,
-      ``,
-      `## Branch`,
-      branch || "(no branch)",
-      ``,
-      `## Uncommitted changes`,
-      diffStat || "(none)",
-      ``,
-      `## Notes`,
-      `Session ended unexpectedly (likely Claude rate/context limit).`,
-      `Resume with: agentdesk team ${taskId}`,
-      ``,
-    ].join("\n");
-    try { writeFileSync(resumePath, resumeContent); } catch {}
-
-    emit({ type: "session:end", duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens, status: "handoff" });
-  } else {
-    // Clean exit — remove stale resume file if present
-    const resumePath = join(cwd, ".agentdesk-resume.md");
-    try { if (existsSync(resumePath)) unlinkSync(resumePath); } catch {}
-
-    emit({ type: "session:end", duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens });
-  }
-
-  sandbox.cleanup();
-
-  return { duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens, handoff: isHandoff };
-}
-
-// --- Phased orchestrator: runs 3 sequential Claude processes ---
-
-async function runSinglePhase({ prompt, cwd, env, teamNames, emit, modelArgs = [], scratchHome, sessionId, tracker, config }) {
-  const wrapped = wrapIsolatedSpawn({
-    cmd: "claude",
-    args: ["-p", prompt, ...modelArgs, "--allowedTools", "Bash,Read,Edit,Write,Glob,Grep", "--verbose", "--output-format", "stream-json"],
-    cwd,
-    scratchHome,
-    sessionId,
-  });
-  const child = spawn(
-    wrapped.cmd,
-    wrapped.args,
-    { stdio: ["pipe", "pipe", "inherit"], shell: false, env, cwd }
-  );
-  child.stdin.end();
-
-  let inputTokens = 0, outputTokens = 0, steps = 0, lastPhase = null;
-
-  const { parseLine } = createStreamParser({
-    teamNames,
-    callbacks: {
-      onPhaseChange({ phase }) { lastPhase = phase; },
-      onModel({ model }) { emit({ type: "session:model", model }); },
-      onAgentMessage({ agent, tag, message }) { emit({ type: "agent:message", agent, tag, message }); },
-      onToolUse({ agent, tool, description }) { steps++; emit({ type: "tool:use", agent, tool, description }); },
-      onToolResult({ success, summary }) { emit({ type: "tool:result", success, summary }); },
-      onSessionUpdate({ taskId: newTaskId, title }) {
-        if (newTaskId) {
-          const update = { type: "session:update", taskId: newTaskId };
-          const taskLink = buildTrackerUrl({ tracker, config, taskId: newTaskId });
-          if (taskLink) update.taskLink = taskLink;
-          emit(update);
-        }
-        if (title) emit({ type: "session:update", title });
-      },
-      onSessionEnd({ inputTokens: iT, outputTokens: oT, steps: s }) {
-        inputTokens = iT; outputTokens = oT; steps = s;
-      },
-    },
-  });
-
-  const rl = createInterface({ input: child.stdout });
-  for await (const line of rl) {
-    parseLine(line);
-  }
-
-  const exitCode = await new Promise(resolve => child.on("close", resolve));
-  return { exitCode, inputTokens, outputTokens, steps, child };
-}
-
-export async function runPhasedOrchestrator({
-  taskId, taskLink, description, createTask, tracker, config,
-  project, team, teamSections, sessionUrl, cwd,
-  onEvent, apiKey, serverUrl, onChild, sessionId,
-}) {
-  const trackerCreds = await fetchTrackerCredentials(project?.name, apiKey, serverUrl);
-
-  // Resolve creds (server + project .env) before any scratch HOME setup
-  // so preflight can fail fast and the sandbox helper bakes the right
-  // token. Server creds win over .env when both are present.
-  const resolvedToken = resolveGitHubCreds({ cwd, trackerCreds }).GITHUB_TOKEN;
-  const resolvedCreds = { ...trackerCreds, GITHUB_TOKEN: resolvedToken || trackerCreds.GITHUB_TOKEN };
-
-  const startTime = Date.now();
-  const teamNames = teamSections.names;
-
-  function timestamp() {
-    const d = new Date();
-    return [d.getHours(), d.getMinutes(), d.getSeconds()].map(n => String(n).padStart(2, "0")).join(":");
-  }
-
-  function emit(event) {
-    onEvent?.({ ...event, timestamp: timestamp() });
-  }
-
-  // Emit session start once
-  emit({
-    type: "session:start",
-    taskId, taskLink,
-    title: description || taskId,
-    project: project?.name || null,
-    sessionNumber: 1,
-    agents: teamNames,
-    cliVersion: CLI_VERSION,
-  });
-
-  // Preflight: AFTER session:start so the UI can bind the error, BEFORE
-  // touching the scratch sandbox or spawning a child. Same contract as
-  // runOrchestrator above.
-  try {
-    assertPushable({ cwd, creds: resolvedCreds, projectName: project?.name });
-  } catch (err) {
-    if (err instanceof PreflightError) {
-      emit({ type: "session:error", code: err.code, message: err.message });
-      emit({ type: "session:end", duration: `${((Date.now() - startTime) / 1000).toFixed(1)}s`, steps: 0, inputTokens: 0, outputTokens: 0, status: "error" });
-      return { duration: "0s", steps: 0, inputTokens: 0, outputTokens: 0, handoff: false, error: err.message };
-    }
-    throw err;
-  }
-
-  const sandbox = createScratchHome({
-    projectId: project?.name,
-    sessionId,
-    creds: resolvedCreds,
-    commitIdentity: {
-      name: config?.identityBadge || "AgentDesk",
-      email: resolvedCreds.JIRA_EMAIL || `agentdesk@local`,
-    },
-  });
-  const env = { ...process.env, ...loadDotEnv(cwd), ...sandbox.env };
-
-  const probe = probeIsolation();
-  if (probe.kind !== "none") {
-    console.error(`[agentdesk] session isolation: ${probe.kind}`);
-  } else {
-    console.error(`[agentdesk] hard isolation unavailable (${probe.reason}) — scoped-env only`);
-  }
-
-  const sessionMemoryPath = join(cwd, ".agentdesk", "session-memory.md");
-  const reviewVerdictPath = join(cwd, ".agentdesk", "review-verdict.md");
-  const MAX_REVIEW_RETRIES = 1; // One execution redo after a failed review, then force SUMMARY.
-  let totalInputTokens = 0, totalOutputTokens = 0, totalSteps = 0;
-  let handoff = false;
-  let reviewRetries = 0;
-
-  // Ordered queue of phases; REVIEW may re-enqueue EXECUTION before SUMMARY.
-  const queue = ["INTAKE", "PLAN", "EXECUTION", "REVIEW", "SUMMARY"];
-
-  while (queue.length > 0) {
-    const phase = queue.shift();
-
-    let sessionMemory = "";
-    try {
-      if (existsSync(sessionMemoryPath)) sessionMemory = readFileSync(sessionMemoryPath, "utf-8").trim();
-    } catch {}
-
-    // Clear any stale verdict before entering REVIEW so we don't read a previous run's result.
-    if (phase === "REVIEW") {
-      try { if (existsSync(reviewVerdictPath)) unlinkSync(reviewVerdictPath); } catch {}
-    }
-
-    emit({ type: "phase:change", phase, model: modelLabelForPhase(phase, config?.phaseModels) });
-
-    const prompt = buildPhasedPrompt({
-      phase, taskId, taskLink, description,
-      createTask: phase === "INTAKE" ? createTask : false,
-      tracker, config, project, teamSections, sessionUrl, cwd, sessionMemory,
-    });
-
-    const modelArgs = modelArgsForPhase(phase, config?.phaseModels);
-    console.error(`[agentdesk] phase ${phase} model: ${modelArgs.length ? modelArgs[1] : "default (no --model flag)"}`);
-    const result = await runSinglePhase({ prompt, cwd, env, teamNames, emit, modelArgs, scratchHome: sandbox.home, sessionId, tracker, config });
-
-    if (onChild) onChild(result.child);
-
-    totalInputTokens += result.inputTokens;
-    totalOutputTokens += result.outputTokens;
-    totalSteps += result.steps;
-
-    const hasMemory = existsSync(sessionMemoryPath);
-
-    if (result.exitCode !== 0 && !hasMemory) {
-      handoff = true;
-
-      let branch = "", diffStat = "";
-      try { branch = execSync("git branch --show-current", { cwd, encoding: "utf-8" }).trim(); } catch {}
-      try { diffStat = execSync("git diff --stat HEAD", { cwd, encoding: "utf-8" }).trim(); } catch {}
-
-      const resumePath = join(cwd, ".agentdesk-resume.md");
-      try {
-        writeFileSync(resumePath, [
-          `# AgentDesk Resume — ${taskId}`,
-          ``, `Session: ${sessionUrl}`, `Phase: ${phase}`,
-          `Date: ${new Date().toISOString()}`,
-          ``, `## Branch`, branch || "(none)",
-          ``, `## Uncommitted changes`, diffStat || "(none)",
-          ``, `## Notes`, `Phased session interrupted during ${phase}.`,
-        ].join("\n"));
-      } catch {}
-      break;
-    }
-
-    // After REVIEW, check the verdict file. NEEDS_MORE_WORK -> loop back to EXECUTION (up to MAX_REVIEW_RETRIES).
-    if (phase === "REVIEW") {
-      let verdict = "";
-      try {
-        if (existsSync(reviewVerdictPath)) verdict = readFileSync(reviewVerdictPath, "utf-8").trim();
-      } catch {}
-
-      const needsMoreWork = /^NEEDS_MORE_WORK\b/i.test(verdict);
-      if (needsMoreWork && reviewRetries < MAX_REVIEW_RETRIES) {
-        reviewRetries++;
-        emit({ type: "agent:message", agent: "Jane", tag: "SAY",
-          message: `Review flagged gaps — returning to EXECUTION (retry ${reviewRetries}/${MAX_REVIEW_RETRIES}).` });
-        queue.unshift("EXECUTION", "REVIEW"); // redo execution, then review again
-      }
-    }
-  }
-
-  // Clean up internal marker files regardless of outcome
-  try { if (existsSync(reviewVerdictPath)) unlinkSync(reviewVerdictPath); } catch {}
-
-  const duration = `${((Date.now() - startTime) / 1000).toFixed(1)}s`;
-
-  if (handoff) {
-    emit({ type: "session:end", duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens, status: "handoff" });
-  } else {
-    // Clean exit — remove stale resume file
-    const resumePath = join(cwd, ".agentdesk-resume.md");
-    try { if (existsSync(resumePath)) unlinkSync(resumePath); } catch {}
-    emit({ type: "session:end", duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens });
-  }
-
-  sandbox.cleanup();
-
-  return { duration, steps: totalSteps, inputTokens: totalInputTokens, outputTokens: totalOutputTokens, handoff };
-}