@deeeed/metamask-harness 0.56.0 → 0.57.0

package/CHANGELOG.md

@@ -2,6 +2,18 @@
 
 ## Unreleased
 
+## 0.57.0 - 2026-09-18
+
+### Changed
+
+- The task and checklist rules the harness shares with Farmslot now come from the Farmslot packages instead of private copies, against `@farmslot/protocol` 0.29.0 and `@farmslot/agent-runtime` 0.11.0 (`@farmslot/recipe-harness` 0.17.0, `@farmslot/expo-recipe` 0.10.0 follow the protocol pin). Which signal statuses are terminal: `isTerminalWorkerSignalStatus`, so `mark` now also protects a `failed` or `done` verdict from a later non-terminal mark, and `status --watch` no longer treats an unknown status as finished. Which box is step N: `enumerateChecklistCheckboxes`, so the `check diff` gate reads the row `mark N` ticks even when an upstream checklist sits under `## Rules` or inside `<details>`. Per-row durations: `deriveChecklistStepDurations`, the same numbers Command Center shows (a task without a recorded start now shows its first row as `0:00` rather than blank). Step names: `checklistStepName`. Which task a checkout is on: `latestTaskDir`, the task written to last by signal or by checklist. File names come from the protocol constants.
+- `status --watch` reads the harness identity from the readiness record (`sandbox.json`) `prepare` writes on every surface; the `artifacts/harness-provenance.json` it looked for is the recipe-cook skill's locked-harness receipt and exists on skill runs only, so the `harness x.y.z` line was blank on every dispatched run.
+
+### Fixed
+
+- `status --watch` no longer reports `no ongoing task detected` while an agent is busy on a long step. The view judged activity only by task-dir writes (`SIGNAL.json`, row marks) and called the task silent after ten minutes, so an investigation step that ran `mm-harness call` every few seconds without ticking a row read as abandoned. The command journal (`temp/recipe/runtime/last-command.json`) now counts as activity, and the view shows it: `last command call ui.navigate · pass · 0:12 ago`. Only the command name and a leading positional argument reach the view; option values never do, and the journal is read through the same hardened reader the harness uses (symlinks and malformed records mean no command line, never a dead watch). A journal older than the task's start belongs to an earlier task and does not count. Piped output logs `command call ui.navigate pass` when the journal moves. `status` itself is not journaled, so watching a checkout never counts as its own activity.
+- `status --watch` renders a Farm checklist the way the Farmslot run view does. Steps, numbers and phases now come from `enumerateChecklistCheckboxes` in `@farmslot/protocol`, the same enumeration the gateway progress view and the `mark` helper use, instead of a private row pattern that rejected Farm's `- [x] **3. Resolve branch:**` rows and only knew `##` headings. A dispatched run therefore showed no rows at all; now it shows `Setup 5/5`, `Reproduce (BEFORE any code changes) 3/9`, … with the bold title as each step's name.
+
 ## 0.56.0 - 2026-09-18
 
 ### Changed

package/dist/commands/checklist.js

@@ -2,6 +2,8 @@ import { spawnSync } from "node:child_process";
 import fs from "node:fs";
 import { createRequire } from "node:module";
 import path from "node:path";
+import { HANDOFF_INPUT } from "@farmslot/agent-runtime";
+import { enumerateChecklistCheckboxes, isTerminalWorkerSignalStatus } from "@farmslot/protocol";
 import { EXIT } from "./shared.js";
 const require2 = createRequire(import.meta.url);
 const MARK_SCRIPT = require2.resolve(
@@ -88,7 +90,7 @@ function runCloseout(taskDir, args, stdio) {
   }
 }
 function stageLearningPackage(taskDir, step, markArgs) {
-  if (!["complete", "no-change"].includes(step ?? "") || markArgs.includes("--skip-learnings") || !fs.existsSync(path.join(taskDir, "inputs", "handoff.json"))) {
+  if (!["complete", "no-change"].includes(step ?? "") || markArgs.includes("--skip-learnings") || !fs.existsSync(path.join(taskDir, HANDOFF_INPUT))) {
     return;
   }
   try {
@@ -108,7 +110,7 @@ function terminalSignalStatus(taskDir) {
   } catch {
     return null;
   }
-  return signal.status === "complete" || signal.status === "blocked" ? signal.status : null;
+  return isTerminalWorkerSignalStatus(signal.status) ? signal.status ?? null : null;
 }
 function preserveTerminalSignal(taskDir, step) {
   if (step !== "start" && numericMarkStep(step) === null) return false;
@@ -124,20 +126,8 @@ function numericMarkStep(step) {
   return Number.isInteger(stepNumber) && stepNumber >= 1 ? stepNumber : null;
 }
 function checklistStepRequiresDiffPass(checklist, stepNumber) {
-  let inFence = false;
-  let seen = 0;
-  for (const line of checklist.split(/\n/u)) {
-    if (/^\s*```/u.test(line)) {
-      inFence = !inFence;
-      continue;
-    }
-    if (inFence) continue;
-    const match = line.match(/^\s*[-*]\s+\[(?: |x|X)\]\s+(.*)$/u);
-    if (!match) continue;
-    seen += 1;
-    if (seen === stepNumber) return /mm-harness check diff/u.test(match[1] ?? "");
-  }
-  return false;
+  const step = enumerateChecklistCheckboxes(checklist).find((item) => item.stepNumber === stepNumber);
+  return step !== void 0 && /mm-harness check diff/u.test(step.rawLabel);
 }
 function checkDiffGateReady(taskDir, step) {
   const stepNumber = numericMarkStep(step);

package/dist/commands/status-watch.js

@@ -1,11 +1,11 @@
 import os from "node:os";
 import path from "node:path";
+import { isTerminalWorkerSignalStatus } from "@farmslot/protocol";
 import { color } from "../cli-color.js";
 import {
   CHECKLIST_LABEL_LIMIT,
   collectTaskView,
   findTaskDir,
-  isTerminalStatus,
   TASK_SILENT_AFTER_MS
 } from "../task-view.js";
 import { EXIT } from "./shared.js";
@@ -59,7 +59,7 @@ async function runStatusWatch(options, deps) {
 `);
     }
     previous = view;
-    const terminal = isTerminalStatus(view.signal?.status);
+    const terminal = isTerminalWorkerSignalStatus(view.signal?.status);
     if (view.activity.ongoing) sawInProgress = true;
     if (!options.watch || terminal && (sawInProgress || options.taskDir !== void 0)) return EXIT.ok;
     await deps.wait(REFRESH_MS);
@@ -106,6 +106,14 @@ function renderTaskView(view, title, explicitTask = false) {
     showTask && view.sandbox?.ready !== void 0 && `sandbox ${view.sandbox.ready ? out("ok", "ready") : out("warn", "not ready")}`
   ]);
   if (contextLine) lines.push(contextLine);
+  const last = view.lastCommand;
+  if (showTask && last) {
+    lines.push(joinParts([
+      `last command ${out("info", joinParts([last.command, last.subject], " "))}`,
+      last.verdict && statusWord(out, last.verdict),
+      last.ageMs !== void 0 && `${formatDuration(last.ageMs)} ago`
+    ]));
+  }
   lines.push("");
   if (!view.activity.ongoing) {
     lines.push(out("warn", "no ongoing task detected"));
@@ -169,6 +177,9 @@ function changedLines(previous, next) {
   if (previous.signal?.status !== next.signal?.status) {
     lines.push(`status ${next.signal?.status ?? "unknown"}`);
   }
+  if (next.lastCommand && previous.lastCommand?.at !== next.lastCommand.at) {
+    lines.push(`command ${joinParts([next.lastCommand.command, next.lastCommand.subject], " ")}${next.lastCommand.verdict ? ` ${next.lastCommand.verdict}` : ""}`);
+  }
   return lines;
 }
 function formatDuration(ms) {

package/dist/task-view.js

@@ -1,34 +1,24 @@
 import { spawnSync } from "node:child_process";
 import fs from "node:fs";
 import path from "node:path";
+import { HANDOFF_INPUT, latestTaskDir } from "@farmslot/agent-runtime";
+import {
+  checklistStepName,
+  deriveChecklistStepDurations,
+  enumerateChecklistCheckboxes,
+  INTERACTIVE_CHECKLIST_MARKDOWN,
+  isTerminalWorkerSignalStatus,
+  READINESS_RECORD,
+  WORKER_SIGNAL_FILE
+} from "@farmslot/protocol";
+import { readCommandJournal } from "./command-journal.js";
 import { resolveRuntimeContextPath } from "./harness.js";
+import { recipeRuntimeDir } from "./paths.js";
 import { parseLibraryPathEnv } from "./review/knowledge.js";
-const TASK_SEARCH_DEPTH = 3;
 const CHECKLIST_LABEL_LIMIT = 70;
-const CHECKLIST_ROW = /^\s*-\s\[([ xX])\]\s*(\d+)\.\s*(.*)$/u;
-const CHECKLIST_PHASE = /^##\s+(.+?)\s*$/u;
 const TASK_SILENT_AFTER_MS = 10 * 60 * 1e3;
 function findTaskDir(target) {
-  const root = path.join(target, "temp", "tasks");
-  let bySignal;
-  let byChecklist;
-  const consider = (best, dir, file) => {
-    const stat = fs.statSync(path.join(dir, file), { throwIfNoEntry: false });
-    if (!stat?.isFile()) return best;
-    return best && best.mtimeMs >= stat.mtimeMs ? best : { dir, mtimeMs: stat.mtimeMs };
-  };
-  const visit = (dir, depth) => {
-    for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
-      if (!entry.isDirectory()) continue;
-      const child = path.join(dir, entry.name);
-      bySignal = consider(bySignal, child, "SIGNAL.json");
-      byChecklist = consider(byChecklist, child, "CHECKLIST.md");
-      if (depth < TASK_SEARCH_DEPTH) visit(child, depth + 1);
-    }
-  };
-  if (!fs.statSync(root, { throwIfNoEntry: false })?.isDirectory()) return void 0;
-  visit(root, 1);
-  return (bySignal ?? byChecklist)?.dir;
+  return latestTaskDir(path.join(target, "temp", "tasks"));
 }
 function collectTaskView(target, taskDir, now) {
   const base = {
@@ -39,20 +29,29 @@ function collectTaskView(target, taskDir, now) {
     )
   };
   if (!taskDir) return { ...base, checklist: [], activity: { ongoing: false, reason: "no-task" } };
-  const handoff = readJsonObject(path.join(taskDir, "inputs", "handoff.json"));
-  const provenance = readJsonObject(path.join(taskDir, "artifacts", "harness-provenance.json"));
+  const handoff = readJsonObject(path.join(taskDir, HANDOFF_INPUT));
   const fixtures = readJsonObject(path.join(taskDir, "artifacts", "prepare", "fixtures.json"));
-  const sandbox = readJsonObject(path.join(taskDir, "artifacts", "sandbox.json"));
-  const signal = readJsonObject(path.join(taskDir, "SIGNAL.json"));
-  const checklist = readChecklistRows(path.join(taskDir, "CHECKLIST.md"));
+  const sandbox = readJsonObject(path.join(taskDir, "artifacts", READINESS_RECORD)) ?? readJsonObject(path.join(target, recipeRuntimeDir(), READINESS_RECORD));
+  const signal = readJsonObject(path.join(taskDir, WORKER_SIGNAL_FILE));
+  const checklist = readChecklistRows(path.join(taskDir, INTERACTIVE_CHECKLIST_MARKDOWN));
   const template = asObject(handoff?.executionTemplate);
   const task = asObject(handoff?.task);
-  const startedAt = parseTime(pickString(handoff, "startedAt"));
-  const timing = applyTiming(checklist, signal, startedAt, now);
+  const startedAtIso = pickString(handoff, "startedAt");
+  const startedAt = parseTime(startedAtIso);
+  const timing = applyTiming(checklist, signal, startedAtIso, now);
   const ticket = pickString(task, "ticket");
+  const journal = readLastCommand(target, now);
+  const commandAt = parseTime(journal?.at);
+  const commandCounts = commandAt !== void 0 && (startedAt === void 0 || commandAt >= startedAt);
+  const lastCommand = commandCounts ? journal : void 0;
   const activity = taskActivity({
     status: pickString(signal, "status"),
-    lastWrites: [startedAt, parseTime(pickString(signal, "timestamp")), ...readMarkTimes(signal).values()],
+    lastWrites: [
+      startedAt,
+      parseTime(pickString(signal, "timestamp")),
+      ...readMarkEvents(signal).map((event) => Date.parse(event.checkedAt)),
+      commandCounts ? commandAt : void 0
+    ],
     branch: base.checkout.branch,
     ticket,
     now
@@ -60,9 +59,10 @@ function collectTaskView(target, taskDir, now) {
   return {
     ...base,
     taskDir,
+    // The readiness record is the only place a harness records itself for a task.
     harness: section({
-      version: pickString(asObject(provenance?.package), "version"),
-      sourceKind: pickString(asObject(provenance?.source), "kind")
+      version: pickString(asObject(sandbox?.harness), "version"),
+      sourceKind: pickString(asObject(sandbox?.harness), "source")
     }),
     template: section({
       id: pickString(template, "id"),
@@ -87,10 +87,26 @@ function collectTaskView(target, taskDir, now) {
       timestamp: pickString(signal, "timestamp"),
       attemptId: pickString(signal, "attemptId")
     }),
+    ...lastCommand ? { lastCommand } : {},
     ...timing,
     activity
   };
 }
+function readLastCommand(target, now) {
+  const { record } = readCommandJournal(target);
+  if (!record) return void 0;
+  const first = record.args[0];
+  const subject = first !== void 0 && !first.startsWith("-") ? first : void 0;
+  const at = record.finishedAt || record.startedAt;
+  const atMs = parseTime(at);
+  return {
+    command: record.command,
+    ...subject ? { subject } : {},
+    verdict: record.verdict,
+    ...at ? { at } : {},
+    ...atMs !== void 0 ? { ageMs: Math.max(0, now - atMs) } : {}
+  };
+}
 function taskActivity(input) {
   const writes = input.lastWrites.filter((value) => value !== void 0);
   const idleMs = writes.length > 0 ? input.now - Math.max(...writes) : void 0;
@@ -100,71 +116,59 @@ function taskActivity(input) {
     ...branchCarriesTicket !== void 0 ? { branchCarriesTicket } : {}
   };
   if (input.status === void 0 && writes.length === 0) return { ongoing: false, reason: "no-task", ...extra };
-  if (isTerminalStatus(input.status)) return { ongoing: false, reason: "finished", ...extra };
+  if (isTerminalWorkerSignalStatus(input.status)) return { ongoing: false, reason: "finished", ...extra };
   if (idleMs !== void 0 && idleMs > TASK_SILENT_AFTER_MS) return { ongoing: false, reason: "silent", ...extra };
   return { ongoing: true, ...extra };
 }
-function isTerminalStatus(status) {
-  return status !== void 0 && status !== "running" && status !== "started";
-}
-function applyTiming(rows, signal, startedAt, now) {
-  const marks = readMarkTimes(signal);
-  let previousAt = startedAt;
-  let lastAt;
-  let firstAt;
+function applyTiming(rows, signal, startedAtIso, now) {
+  const events = readMarkEvents(signal);
+  const durations = new Map(
+    deriveChecklistStepDurations({ schemaVersion: 1, events }, startedAtIso).map((step) => [
+      step.stepNumber,
+      step.durationMs
+    ])
+  );
+  const marks = new Map(events.map((event) => [event.stepNumber, Date.parse(event.checkedAt)]));
   for (const row of rows) {
-    const checkedAt = marks.get(row.number);
-    if (checkedAt === void 0) continue;
-    if (previousAt !== void 0) row.durationMs = checkedAt - previousAt;
-    if (firstAt === void 0) firstAt = checkedAt;
-    previousAt = checkedAt;
-    lastAt = checkedAt;
+    const durationMs = durations.get(row.number);
+    if (durationMs !== void 0) row.durationMs = durationMs;
   }
+  const instants = [...marks.values()];
+  const lastAt = instants.length > 0 ? Math.max(...instants) : void 0;
+  const firstAt = instants.length > 0 ? Math.min(...instants) : void 0;
   const current = rows.find((row) => !row.checked);
   if (current && lastAt !== void 0) current.runningMs = now - lastAt;
-  const origin = startedAt ?? firstAt;
+  const origin = parseTime(startedAtIso) ?? firstAt;
   return {
     ...current ? { currentStep: current.number } : {},
     ...origin !== void 0 ? { elapsedMs: now - origin } : {}
   };
 }
-function readMarkTimes(signal) {
-  const marks = /* @__PURE__ */ new Map();
+function readMarkEvents(signal) {
   const events = asObject(signal?.checklistTiming)?.events;
-  if (!Array.isArray(events)) return marks;
+  if (!Array.isArray(events)) return [];
+  const marks = [];
   for (const entry of events) {
     const event = asObject(entry);
     const stepNumber = pickNumber(event, "stepNumber");
-    const checkedAt = parseTime(pickString(event, "checkedAt"));
-    if (stepNumber === void 0 || checkedAt === void 0) continue;
-    marks.set(stepNumber, checkedAt);
+    const checkedAt = pickString(event, "checkedAt");
+    if (stepNumber === void 0 || parseTime(checkedAt) === void 0) continue;
+    marks.push({ stepNumber, label: "", checkedAt });
   }
   return marks;
 }
 function readChecklistRows(file) {
   const text = readTextFile(file);
   if (text === void 0) return [];
-  const rows = [];
-  let phase;
-  for (const line of text.split("\n")) {
-    const heading = CHECKLIST_PHASE.exec(line);
-    if (heading) {
-      phase = heading[1];
-      continue;
-    }
-    const match = CHECKLIST_ROW.exec(line);
-    if (!match) continue;
-    rows.push({
-      ...phase ? { phase } : {},
-      number: Number(match[2]),
-      label: trimLabel(match[3]),
-      checked: match[1].toLowerCase() === "x"
-    });
-  }
-  return rows;
+  return enumerateChecklistCheckboxes(text).map((item) => ({
+    ...item.phase ? { phase: item.phase } : {},
+    number: item.stepNumber,
+    label: trimLabel(item.rawLabel),
+    checked: item.checked
+  }));
 }
 function trimLabel(text) {
-  const plain = text.replace(/\*\*/gu, "").replace(/\s+/gu, " ").trim();
+  const plain = checklistStepName(text).replace(/\*\*/gu, "").replace(/^\d+\.\s*/u, "").replace(/\s+/gu, " ").trim();
   if (plain.length <= CHECKLIST_LABEL_LIMIT) return plain;
   return `${plain.slice(0, CHECKLIST_LABEL_LIMIT).trimEnd()}\u2026`;
 }
@@ -242,6 +246,5 @@ export {
   CHECKLIST_LABEL_LIMIT,
   TASK_SILENT_AFTER_MS,
   collectTaskView,
-  findTaskDir,
-  isTerminalStatus
+  findTaskDir
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@deeeed/metamask-harness",
-  "version": "0.56.0",
+  "version": "0.57.0",
   "type": "module",
   "bin": {
     "mm-harness": "bin/mm-harness"
@@ -20,11 +20,11 @@
     "check:syntax": "find . -name '*.mjs' -print0 | xargs -0 -n1 node --check"
   },
   "dependencies": {
-    "@farmslot/agent-runtime": "0.10.0",
-    "@farmslot/expo-recipe": "0.9.2",
+    "@farmslot/agent-runtime": "0.11.0",
+    "@farmslot/expo-recipe": "0.10.0",
     "@farmslot/handoff": "^0.3.1",
-    "@farmslot/protocol": "0.28.0",
-    "@farmslot/recipe-harness": "0.16.3",
+    "@farmslot/protocol": "0.29.0",
+    "@farmslot/recipe-harness": "0.17.0",
     "agent-device": "0.19.3",
     "commander": "^12.0.0",
     "es-module-lexer": "2.3.1",