@deeeed/metamask-harness 0.53.0 → 0.54.1

package/CHANGELOG.md

@@ -2,6 +2,18 @@
 
 ## Unreleased
 
+## 0.54.1 - 2026-09-18
+
+### Changed
+
+- `help --json` quality section carries the verdict rules (pass, warn, fail), the shape of an independent critique, and per-adapter review notes for Mobile, Extension and Core; the separate recipe-quality skill is retired in favour of this guidance.
+
+## 0.54.0 - 2026-09-17
+
+### Changed
+
+- Farmslot protocol 0.28.0 and agent-runtime 0.10.0: an execution template no longer declares a run mode, so `task init` stops defaulting `--run-mode`. The flag stays accepted for a control plane that has one to pass; omit it for a skill or manual run and nothing invents a mode. Help, README, and site examples use the plain template ids the catalog now offers (`fix-bug/mobile`, `review-pr/static-perps.mobile`). Pins recipe-harness 0.16.3 and expo-recipe 0.9.2, the builds that also carry protocol 0.28.0, so a packed install keeps one protocol copy.
+
 ## 0.53.0 - 2026-09-17
 
 ### Added

package/README.md

@@ -129,7 +129,7 @@ mm-harness execution-template list --package-templates <catalog> --json
 
 # Write the task directory the agent works from.
 mm-harness task init temp/tasks/fix/TAT-1 \
-  --flow fix-bug --run-mode autonomous --template fix-bug/autonomous.mobile \
+  --flow fix-bug --template fix-bug/mobile \
   --package-templates <catalog> --title "Wrong total on the receipt" --ticket TAT-1
 
 # Bring the checkout to a proven-ready state before the first step.
@@ -271,7 +271,7 @@ Unset `MM_HARNESS_BIN` to return to the published installation.
 
 ### Workflow entry points and completion
 
-Explicitly invoke `mms-recipe-dev`, `mms-recipe-fixbug`, `mms-recipe-qa` or the team's static review skill. `mms-recipe-cook` selects or resumes their shared Markdown checklist. Matching descriptions or changed paths never activates these workflows.
+Explicitly invoke `mms-recipe-dev`, `mms-recipe-fix-bug`, `mms-recipe-qa` or the team's static review skill. `mms-recipe-cook` selects or resumes their shared Markdown checklist. Matching descriptions or changed paths never activates these workflows.
 
 Static Perps review uses a generated source checklist without installing or launching the app harness. `recipe-qa` freezes a PR, range or daily change window and requires runtime smoke plus selected recipe proof. Official release validation uses its named release profile and proof lane.
 

package/dist/command-contract.js

@@ -158,7 +158,7 @@ const PUBLIC_COMMAND_CONTRACTS = {
     })
   },
   task: {
-    usage: "mm-harness task init <task-dir> --flow f --run-mode m --template id --title t [options]",
+    usage: "mm-harness task init <task-dir> --flow f --template id --title t [options]",
     options: options(HELP, JSON, {
       "--flow": value(),
       "--run-mode": value(["autonomous", "interactive", "validation"]),

package/dist/commands/task-init.js

@@ -11,7 +11,6 @@ function hasOption(argv, name) {
 function metamaskDefaults(argv) {
   const defaults = [];
   if (!hasOption(argv, "--surface")) defaults.push("--surface", "skill");
-  if (!hasOption(argv, "--run-mode")) defaults.push("--run-mode", "autonomous");
   if (!hasOption(argv, "--mark-command")) {
     defaults.push("--mark-command", `${harnessExecutable()} checklist mark`);
   }

package/dist/mm-harness-cli.js

@@ -275,7 +275,7 @@ Example:
   {
     name: "task",
     summary: "Create the task directory an agent works from, out of a shared execution template.",
-    example: 'mm-harness task init temp/tasks/fix/TAT-1 --flow fix-bug --run-mode autonomous --template fix-bug/autonomous.mobile --title "Wrong total"',
+    example: 'mm-harness task init temp/tasks/fix/TAT-1 --flow fix-bug --template fix-bug/mobile --title "Wrong total"',
     helpText: `mm-harness task init <task-dir> [options]
 
   Write the task directory from a selected execution template: TASK.md, the
@@ -283,8 +283,8 @@ Example:
   mm-harness checklist mark, and inputs/ (handoff.json, worker-terminal-contract.json).
   checklist-target.json is NOT written \u2014 absent means CHECKLIST.md + SIGNAL.json.
 
-  Applied when you do not pass them: --surface skill, --run-mode autonomous,
-  --mark-command for this executable, and --platform from the detected checkout.
+  Applied when you do not pass them: --surface skill, --mark-command for this
+  executable, and --platform from the detected checkout.
 
 Task:
   --title <t>                        Task title (required)
@@ -294,7 +294,9 @@ Task:
 
 Selection (same catalog as execution-template):
   --flow <flow>                      Exact flow (required)
-  --run-mode <mode>                  autonomous|interactive|validation (default autonomous; set by a control plane)
+  --run-mode <mode>                  autonomous|interactive|validation (optional;
+                                     set by the Farmslot control plane, omit for
+                                     skill/manual runs)
   --platform <platform>              Exact platform
   --template <id>                    Exact template id (required)
   --domain <domain>                  Exact domain
@@ -316,8 +318,8 @@ Identity and rendering:
   --json                             Machine-readable output
 
 Example:
-  mm-harness task init temp/tasks/fix/TAT-1 --flow fix-bug --run-mode autonomous \\
-    --template fix-bug/autonomous.mobile --package-templates <path> \\
+  mm-harness task init temp/tasks/fix/TAT-1 --flow fix-bug \\
+    --template fix-bug/mobile --package-templates <path> \\
     --title "Wrong total on the receipt" --ticket TAT-1 --json`
   },
   {
@@ -344,14 +346,16 @@ Common source/filter options:
   --project-name <name>              Source label for project templates
   --flow <flow>                      Exact flow
   --platform <platform>              Exact platform
-  --run-mode <mode>                  autonomous|interactive|validation
+  --run-mode <mode>                  autonomous|interactive|validation (optional;
+                                     set by the Farmslot control plane, omit for
+                                     skill/manual runs)
   --domain <domain>                  Exact domain
   --id <id>                          Exact template id
   --json                             Machine-readable output
 
 Materialize:
   mm-harness execution-template materialize <output> --flow fix-bug \\
-    --platform mobile --run-mode autonomous --id fix-bug/autonomous.mobile \\
+    --platform mobile --id fix-bug/mobile \\
     --package-templates <path> --provenance <path> --json`
   },
   {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@deeeed/metamask-harness",
-  "version": "0.53.0",
+  "version": "0.54.1",
   "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": "0.9.0",
-    "@farmslot/expo-recipe": "0.9.1",
+    "@farmslot/agent-runtime": "0.10.0",
+    "@farmslot/expo-recipe": "0.9.2",
     "@farmslot/handoff": "^0.3.1",
-    "@farmslot/protocol": "0.26.0",
-    "@farmslot/recipe-harness": "0.16.2",
+    "@farmslot/protocol": "0.28.0",
+    "@farmslot/recipe-harness": "0.16.3",
     "agent-device": "0.19.3",
     "commander": "^12.0.0",
     "es-module-lexer": "2.3.1",

package/scripts/site-contrast.mjs

@@ -394,9 +394,9 @@ async function checkBehavior(browserEndpoint, port, pages) {
       ['/mms-recipe-qa', copied.work.join('\n')],
       ['/mms-recipe-cook', copied.work.join('\n')],
       ['/mms-recipe-performance', copied.work.join('\n')],
-      ['/mms-recipe-evidence', copied.work.join('\n')],
-      ['agentic/recipe-performance', copied.setup],
-      ['agentic/recipe-evidence', copied.setup],
+      ['/mms-recipe-pretty-pr', copied.work.join('\n')],
+      ['perps/recipe-performance', copied.setup],
+      ['agentic/recipe-pretty-pr', copied.setup],
     ];
     const missing = expected.filter(([line, text]) => !text?.includes(line)).map(([line]) => line);
     if (missing.length) {
@@ -448,7 +448,7 @@ async function checkBehavior(browserEndpoint, port, pages) {
         return checked;
       })()
     `);
-    if (!previews.includes('review-pr/static-perps') || !previews.includes('qa/pr.extension')) {
+    if (!previews.includes('review-pr/static-perps.mobile') || !previews.includes('qa/pr.extension')) {
       throw new Error('Static review and QA previews must both be available');
     }
     const progress = await page.evaluate(String.raw`

package/site/assets/help-recipes.json

@@ -212,6 +212,26 @@
             "mobile",
             "extension"
           ]
+        },
+        "Verdict: pass when every named claim ran an executable path with an assertion and consistent evidence; warn when the supported claims pass but non-blocking gaps remain; fail when a named claim has no executable path or assertion, the graph can pass unconditionally, required evidence is missing or contradictory, or the relevant path did not run. A visual claim cannot pass without a settled capture. Write only the compact {verdict, reasons} to artifacts/recipe-quality-input.json and build the artifact with mm-harness recipe-quality build; never hand-author it.",
+        "Independent critique, when one is requested: the verdict with one evidence-backed sentence, uncovered or weak proof targets, graph, composition, adapter and reuse findings, evidence mismatches and flake risks, and the three highest-value concrete edits.",
+        {
+          "text": "State the simulator or device, build type and app state. Unlock through a fixture or a documented primitive; keep account, balance, network, permissions and feature flags deterministic; use existing page objects, test ids and fixtures; capture only after ui.wait_for, a route assertion or a state assertion; teardown must not leak balances, pending transactions, selected network, permissions or onboarding state into the next run.",
+          "adapters": [
+            "mobile"
+          ]
+        },
+        {
+          "text": "State browser, build, fixture and the popup, fullscreen or service-worker context. Popup UI proven only through CDP eval or controller reads fails a visual claim; capture after settle; prefer existing e2e and fixture helpers over raw steps.",
+          "adapters": [
+            "extension"
+          ]
+        },
+        {
+          "text": "Core is headless: prove claims with summary.json, trace.json, logs or package tests, never screenshots; use manifest-declared headless actions only; state the target package or workspace, fixture, start state and teardown for controller mutations; a non-visual criterion needs unit or package test coverage.",
+          "adapters": [
+            "core"
+          ]
         }
       ],
       "commands": [