@shipster/cli 0.5.0 → 0.5.1

package/dist/event-worker.js

@@ -263,7 +263,8 @@ You implement specs on a branch and open a PR. Write production code and tests.
 
 **Playbook rules:**
 - Take the PR head sha from the PR itself (\`gh pr view <n> --json headRefOid\`), never from a handoff message or a task comment.
-- One branch per order, cut from the default branch: never stack on another order's unmerged branch.`;
+- One branch per order, cut from the default branch: never stack on another order's unmerged branch.
+- A test may delete only paths under a directory it created with mkdtemp in that run. Never remove, empty or clear /tmp, $TMPDIR or any other shared directory as a whole, and never point a cleanup or reset at the real one.`;
     case "tester":
       return `## Your role: Tester
 
@@ -277,7 +278,8 @@ Write, fix, and run tests on the branch in branchName metadata.
 - If tests cannot pass due to an implementation bug: \`mcp__relai__add_task_comment\`
   (type: "finding") with type "bug", severity "blocking", then
   \`mcp__relai__assign_task\` back to the writer (previous assignee or list_agents).
-  Do not complete.`;
+  Do not complete.
+- A test may delete only paths under a directory it created with mkdtemp in that run. Never remove, empty or clear /tmp, $TMPDIR or any other shared directory as a whole, and never point a cleanup or reset at the real one.`;
     case "devops":
       return `## Your role: DevOps
 
@@ -293,6 +295,7 @@ Handle CI/CD, infrastructure, build/deploy configuration.
       throw new Error(`Unknown specialization "${spec}" \u2014 valid values: reviewer, architect, writer, tester, devops`);
   }
 }
+var PR_FROM_ORDER = "In this section `<prUrl>` is the `prUrl` that `mcp__relai__get_order` returns for `metadata.orderId`. Never use the task's own `metadata.prUrl`. If the `epicId` get_order returns is not this task's own `epicId`, if get_order has no `prUrl`, or the task's `metadata.prUrl` is set and differs from it, run no `gh` command: block the task with a question naming the orchestrator, and stop.";
 var ORDER_RULES = `## Factory orders
 
 An order's tasks carry
@@ -361,7 +364,7 @@ Find the default branch with \`git remote show origin\`.
      is done. With one gate, block it on the build tasks alone.
    - \`ci_passed\` gets no gate task: the merge task checks CI. Say so on the planning task's thread.
 5. One merge task, blocked by every gate task (or every build task when there are no gates),
-   \`assignedTo: "@auto"\`, \`metadata: { orderId, orderRole: "merge", prUrl }\`.
+   \`assignedTo: "@auto"\`, \`metadata: { orderId, orderRole: "merge" }\`. The server stamps \`prUrl\` on it from the order, so you may omit it.
    If the order has \`pairBuilds\`, every build task needs \`verifyKind: "reviewer_agent"\` and a
    \`verifyReviewerId\` on a different \`workerType\` from the writers (see \`mcp__relai__list_agents\`);
    the API refuses a paired build task without one and routes it away from the reviewer's vendor.
@@ -374,26 +377,28 @@ Find the default branch with \`git remote show origin\`.
 \`metadata.orderKind: "review"\`): the goal is one review posted on the existing pull request at the
 order's \`prUrl\`. There is no order branch, no draft PR and no merge, so skip steps 2 to 6 above.
 Nobody pushes to the PR's branch, approves it, or requests changes on it.
+${PR_FROM_ORDER}
 1. As the planner, read the PR with \`gh pr view <prUrl>\` and \`gh pr diff <prUrl>\`. Pick the
    angles it needs: correctness, security, simplicity, tests and consistency with the codebase for a
    real change; correctness and consistency alone for a small, low-risk one (docs, tests, under about
    150 changed lines with no auth, input handling, secrets, money or migrations).
 2. One task per angle, \`epicId\` = the order's \`epicId\`,
-   \`metadata: { orderId, orderRole: "review", angle: "<angle>", prUrl }\`. Spread the angles
+   \`metadata: { orderId, orderRole: "review", angle: "<angle>" }\`; the server stamps \`prUrl\`. Spread the angles
    across vendors: assign each to a different agent from \`mcp__relai__list_agents\`, preferring a
    different \`workerType\` from the agents already given one, and reuse agents only when there are
    more angles than agents. Each description names the angle, the PR, what that angle hunts for, and
    the reporting format below.
 3. One lead task, \`blockedBy\` every angle task, assigned to yourself, with
-   \`metadata: { orderId, orderRole: "review_lead", prUrl }\`.
+   \`metadata: { orderId, orderRole: "review_lead" }\`.
 4. When acceptance includes \`human_approval\`, add a gate task blocked by the lead
    (\`metadata: { orderId, orderRole: "gate" }\`, \`verifyKind: "human_approval"\`) and a post task
-   blocked by the gate (\`metadata: { orderId, orderRole: "post_review", prUrl }\`).
+   blocked by the gate (\`metadata: { orderId, orderRole: "post_review" }\`).
 5. If the order has \`relatedRepos\`, copy them into every angle and lead task's metadata and name them in
    each description, so reviewers check the contracts this PR depends on in those repos.
 6. Post the task list on the planning task's thread and complete it.
 
 **orderRole "review"** (one angle of a review goal):
+- ${PR_FROM_ORDER} For the PR head, \`<n>\` is that PR's number.
 - Read the diff and as much surrounding code as you need. To run anything, check out the PR head
   detached: \`git fetch origin pull/<n>/head && git checkout --detach FETCH_HEAD\`. Never push, never
   comment on the PR, never edit the repo's history.
@@ -409,6 +414,7 @@ Nobody pushes to the PR's branch, approves it, or requests changes on it.
 - Post the report with \`mcp__relai__add_task_comment\` on this task, then complete it.
 
 **orderRole "review_lead"**:
+- ${PR_FROM_ORDER}
 - Read every angle task's report with \`mcp__relai__get_task_comments\`. Merge duplicates, drop
   findings you cannot reproduce or whose premise is wrong, and rank what is left. Keep an UNPROVEN mark
   on any finding about a related repo nobody could read. Severity sets the
@@ -421,7 +427,7 @@ Nobody pushes to the PR's branch, approves it, or requests changes on it.
   and put the review URL on the task's thread. With a gate, post the review as a comment on this task
   for the owner to approve, and do not touch the PR.
 
-**orderRole "post_review"**: once the gate is approved, post the lead's review comment from the
+**orderRole "post_review"**: ${PR_FROM_ORDER} Once the gate is approved, post the lead's review comment from the
 lead task's thread with \`gh pr review <prUrl> --comment --body-file <file>\`, put the review URL on
 this task's thread, and complete.
 
@@ -451,6 +457,8 @@ this task's thread, and complete.
   the rest; if it is not recorded, reject and name it so the builder records it.
 
 **orderRole "merge"**:
+${PR_FROM_ORDER}
+
 **Playbook rules:**
 - Report a merge you cannot clear under step 3 the same minute: post which check failed, at which sha, and who must act on the merge task's thread, then set the task blocked.
 - Never merge on a stale head: re-read the PR head sha immediately before merging, pass it as \`--match-head-commit\`, and stop if it differs from the sha the gates approved, or the carried head after an Exit 0 conflict-only remerge.
@@ -2050,7 +2058,7 @@ function claudeArgs(opts) {
     ...opts.isolateUserConfig ? ["--setting-sources", "project,local"] : []
   ];
 }
-async function runClaudeSession(config2) {
+async function runClaudeSession(config2, opts = {}) {
   const repoContextOn = readEnv("REPO_CONTEXT") === "1";
   const lessonsPromise = repoContextOn ? Promise.resolve("") : lessonsBlock({ ...config2, repoId: taskRepoId(config2) });
   const repoContextPromise = repoContextOn ? (readEnv("TASK_REPO_ID")?.trim() ? Promise.resolve(taskRepoId(config2)) : contextRepoId(config2)).then((repoId) => repoContextBlock({ repoId, repoPath: config2.repoPath, role: config2.specialization }, config2)) : Promise.resolve("");
@@ -2075,7 +2083,8 @@ async function runClaudeSession(config2) {
       const proc = spawn(config2.claudeBin, args, {
         cwd: config2.repoPath,
         env: { ...childEnv, PATH: process.env.PATH },
-        stdio: ["pipe", "pipe", "pipe"]
+        stdio: ["pipe", "pipe", "pipe"],
+        signal: opts.signal
       });
       console.log(`[claude-worker] Process PID: ${proc.pid}`);
       Promise.all([lessonsPromise, repoContextPromise]).then(([lessons, repoContext]) => {
@@ -2234,7 +2243,7 @@ function readCursorEvent(line) {
   }
   return {};
 }
-async function runCursorSession(config2) {
+async function runCursorSession(config2, opts = {}) {
   const model = pinnedModel();
   recordLaunchedModel(model);
   const args = cursorArgs(config2.repoPath, buildPrompt(config2, { lessons: await lessonsBlock({ ...config2, repoId: taskRepoId(config2) }) }), model);
@@ -2242,7 +2251,7 @@ async function runCursorSession(config2) {
   const startedAt = (/* @__PURE__ */ new Date()).toISOString();
   const taskIdsPromise = listSessionTaskIds(config2, "[cursor-worker]");
   return new Promise((resolve2, reject) => {
-    const proc = spawn2(bin(), args, { cwd: config2.repoPath, env: workerEnv("cursor", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"] });
+    const proc = spawn2(bin(), args, { cwd: config2.repoPath, env: workerEnv("cursor", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"], signal: opts.signal });
     console.log(`[cursor-worker] Process PID: ${proc.pid}`);
     const toolsUsed = [];
     let buffer = "";
@@ -2354,7 +2363,7 @@ function readCopilotEvent(line) {
   if (type === "session.warning" && POLICY_BLOCKS.some((re) => re.test(message))) return { error: message };
   return {};
 }
-async function runCopilotSession(config2) {
+async function runCopilotSession(config2, opts = {}) {
   const mcpConfigPath = writeMcpConfig(config2, "copilot-worker");
   const model = pinnedModel();
   recordLaunchedModel(model);
@@ -2363,7 +2372,7 @@ async function runCopilotSession(config2) {
   const taskIdsPromise = listSessionTaskIds(config2, "[copilot-worker]");
   console.log(`[copilot-worker] Spawning: ${bin2()} -p <prompt> ${args.slice(2, 6).join(" ")} ...`);
   return new Promise((resolve2, reject) => {
-    const proc = spawn3(bin2(), args, { cwd: config2.repoPath, env: workerEnv("copilot", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"] });
+    const proc = spawn3(bin2(), args, { cwd: config2.repoPath, env: workerEnv("copilot", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"], signal: opts.signal });
     console.log(`[copilot-worker] Process PID: ${proc.pid}`);
     const toolsUsed = [];
     let buffer = "";
@@ -2481,7 +2490,7 @@ function readCodexEvent(line) {
   if (event.type === "error") return { error: String(event.message ?? "codex error") };
   return {};
 }
-async function runCodexSession(config2) {
+async function runCodexSession(config2, opts = {}) {
   const profile = `pitboss-${config2.agentId}`;
   mkdirSync3(codexHome(), { recursive: true, mode: 448 });
   const profilePath = join5(codexHome(), `${profile}.config.toml`);
@@ -2507,7 +2516,7 @@ async function runCodexSession(config2) {
   const taskIdsPromise = listSessionTaskIds(config2, "[codex-worker]");
   console.log(`[codex-worker] Spawning: ${bin3()} ${args.slice(0, 6).join(" ")} ...`);
   return new Promise((resolve2, reject) => {
-    const proc = spawn4(bin3(), args, { cwd: config2.repoPath, env: workerEnv("codex", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"] });
+    const proc = spawn4(bin3(), args, { cwd: config2.repoPath, env: workerEnv("codex", process.env, grantFromEnv()), stdio: ["ignore", "pipe", "pipe"], signal: opts.signal });
     console.log(`[codex-worker] Process PID: ${proc.pid}`);
     const toolsUsed = [];
     let buffer = "";
@@ -3690,12 +3699,18 @@ function endSession(file) {
     }
   }
 }
+function endsTask(event, taskId) {
+  const task = event?.payload?.task;
+  if (!task || task.id !== taskId) return false;
+  return task.status === "cancelled" || task.archivedAt !== void 0 && task.archivedAt !== null;
+}
 async function runEventWorker(config2) {
   console.log(`[event-worker] Starting \u2014 agent ${config2.agentId}, watching ${config2.apiUrl}/events`);
   console.log(`[event-worker] Repo: ${config2.repoPath} | Runtime: ${config2.runtime} | Model: ${config2.model}`);
   await assertRepoOrExit(config2, "[event-worker]");
   await assertNotOwnerScopedOrExit(config2, "[event-worker]");
   await selfSubscribe(config2);
+  let live = null;
   const queue = createRunQueue(async () => {
     await heartbeat(config2, "[event-worker]");
     const slotMode = readEnv("OWNER_POOL_SLOT") === "1";
@@ -3749,6 +3764,7 @@ async function runEventWorker(config2) {
         return;
       }
     }
+    const controller = new AbortController();
     try {
       console.log("[event-worker] Event received \u2014 running session...");
       const activityFile = readEnv("ACTIVITY_FILE");
@@ -3767,9 +3783,17 @@ async function runEventWorker(config2) {
         setEnv(TASK_REPO_ID_ENV, checkout.repoId);
         setEnv(TASK_REPO_URL_ENV, checkout.repoUrl);
       }
+      if (checkout) live = { taskId: checkout.taskId, controller };
       try {
-        await getAdapter(config2.runtime).runSession(config2);
+        await getAdapter(config2.runtime).runSession(config2, { signal: controller.signal });
       } finally {
+        live = null;
+        if (controller.signal.aborted && readEnv(RESET_SEAT_DIR_ENV)) {
+          try {
+            await killSeatProcesses();
+          } catch {
+          }
+        }
         if (slotMode) envRestore(priorPin);
         if (checkout) {
           envRestore(priorTaskRepo.id);
@@ -3785,6 +3809,10 @@ async function runEventWorker(config2) {
       await clearUnavailable(config2, "[event-worker]");
     } catch (err) {
       const text = err instanceof Error ? err.message : String(err);
+      if (controller.signal.aborted) {
+        console.log(`[event-worker] Session stopped because its task was cancelled or archived (${text.split("\n")[0].slice(0, 120)})`);
+        return;
+      }
       const exitActivityFile = readEnv("ACTIVITY_FILE");
       if (exitActivityFile) {
         try {
@@ -3829,6 +3857,10 @@ async function runEventWorker(config2) {
       try {
         const event = JSON.parse(raw.data);
         console.log(`[event-worker] Event: ${event.kind ?? "unknown"}`);
+        if (live && endsTask(event, live.taskId) && !live.controller.signal.aborted) {
+          console.warn(`[event-worker] task ${live.taskId} was cancelled or archived \u2014 stopping its session`);
+          live.controller.abort();
+        }
       } catch {
       }
       queue.notify();

package/dist/mcp.js

@@ -12925,7 +12925,7 @@ function buildOrderReadTools(client, repoId, opts) {
   return [
     {
       name: "get_order",
-      description: "Read one factory order by id. Status is derived from the Epic's child tasks and acceptance gates, not stored transitions. Also returns `tasks` (id, title, status, assignee, orderRole, blockedBy, model, cost), `gates` (kind, status, reviewer, decision), `cost` ({ totalUsd, totalTurns, byModel, tasksReporting, tasksTotal, note }) and `models`. The cost covers only what workers have reported so far: read its `note` before quoting the total. " + ORDER_GATE_NOTE,
+      description: "Read one factory order by id. Status is derived from the Epic's child tasks and acceptance gates, not stored transitions. Also returns `tasks` (id, title, status, assignee, orderRole, blockedBy, model, cost), `gates` (kind, status, reviewer, decision), `cost` ({ totalUsd, totalTurns, byModel, byTier, tasksReporting, tasksTotal, note }) and `models`. The cost covers only what workers have reported so far: read its `note` before quoting the total. " + ORDER_GATE_NOTE,
       inputSchema: external_exports.object({
         id: external_exports.string().describe("The order id (ord_\u2026).")
       }),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@shipster/cli",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "private": false,
   "description": "Connect your coding agents to a pitboss instance",
   "keywords": [