@deeeed/metamask-harness 0.86.2 → 0.88.0

package/library/manifests/extension.action-manifest.json

@@ -3788,7 +3788,7 @@
       ]
     },
     "metamask.perps.read_snapshot": {
-      "description": "Read a bounded, redacted source-dev snapshot of Extension Perps state through the optional product background hook.",
+      "description": "Read a bounded, redacted source-dev snapshot of Extension Perps state through the optional product background hook. The history section is the funds ledger from perpsGetUserHistory (deposits, withdrawals and transfers): it holds no trades or orders, so zero history items does not mean no trading history. Read trade and order rows with metamask.perps.read_visible_state activity.",
       "schema": {
         "type": "object",
         "properties": {
@@ -3830,6 +3830,7 @@
           },
           "history": {
             "type": "object",
+            "description": "Funds-ledger window (deposits, withdrawals and transfers) passed to perpsGetUserHistory; trades and orders are not part of it.",
             "properties": {
               "start_time_ms": {
                 "type": "integer",
@@ -4131,7 +4132,7 @@
       "execution_capabilities": []
     },
     "metamask.perps.close_orders": {
-      "description": "extension Cancel selected live Perps open orders. Use selector/mode params to cancel matching symbols or all orders. On mainnet, cancel is refused in the fixture-policy lane and needs the funded release lane. Extension ensure_* only assert existing state; converge an owned testnet account through an explicitly funded leaf action.",
+      "description": "extension Cancel selected live Perps open orders. Use selector/mode params to cancel matching symbols or all orders. On mainnet, the fixture-policy lane cancels only orders this run placed; any other cancel needs the funded release lane. Extension ensure_* only assert existing state; converge an owned testnet account through an explicitly funded leaf action.",
       "schema": {
         "type": "object",
         "properties": {
@@ -4252,7 +4253,7 @@
           },
           "order_id": {
             "type": "string",
-            "description": "Exact product order ID required for receipt-bound background cancellation."
+            "description": "Exact product order ID required for receipt-bound background cancellation. On mainnet without funding files, only an order this run placed."
           },
           "offset_pct": {
             "type": "number"
@@ -4409,7 +4410,7 @@
           },
           "close_immediately": {
             "type": "boolean",
-            "description": "Close the opened position again inside this action and prove the market flat (market orders only; the market must start flat). Required on mainnet, with leverage 1 and notional at most $15 from a fixture account."
+            "description": "Close the opened position again inside this action and prove the market flat (market orders only; the market must start flat). A mainnet market order needs it, with leverage 1 and notional at most $15 from a fixture account; a mainnet limit order takes none (the run cancels it at its end)."
           },
           "provider": {
             "type": "string",
@@ -4421,7 +4422,7 @@
               "mainnet",
               "testnet"
             ],
-            "description": "Exact Perps network required before background dispatch; the action never toggles it."
+            "description": "Exact Perps network required before background dispatch; the action never toggles it (metamask.perps.set_network does). A mainnet order is a market order with close_immediately, or a dev1 limit order away from the market at 1x and at most $15 notional at its limit price, which the run cancels at its end (success, failure or abort) and records in perps-mainnet-orders.json."
           },
           "order_type": {
             "type": "string",
@@ -5160,6 +5161,36 @@
         "app-mutation"
       ]
     },
+    "metamask.perps.set_network": {
+      "description": "Switch the Extension's Perps network through the background perpsToggleTestnet request and wait until the app reports it. The fixture loads testnet; network=mainnet is the explicit opt-in and lasts one run: the run's teardown switches back to testnet (success, failure or abort).",
+      "schema": {
+        "type": "object",
+        "properties": {
+          "network": {
+            "type": "string",
+            "enum": ["testnet", "mainnet"],
+            "default": "testnet",
+            "description": "Perps network to select; unchanged when already selected."
+          },
+          "timeout_ms": {
+            "type": "number",
+            "description": "How long to wait for the app to report the new network."
+          }
+        },
+        "additionalProperties": false
+      },
+      "examples": [
+        {
+          "action": "metamask.perps.set_network",
+          "network": "mainnet",
+          "intent": "Switch Perps to mainnet for this run",
+          "next": "done"
+        }
+      ],
+      "execution_capabilities": [
+        "app-mutation"
+      ]
+    },
     "metamask.deeplink.open": {
       "description": "Open an HTTPS deeplink through real browser ingress so the Extension owns parsing, verification, and destination navigation.",
       "execution_capabilities": [

package/library/manifests/mobile.action-manifest.json

@@ -930,7 +930,7 @@
       ]
     },
     "ui.wait_for": {
-      "description": "Wait for a React Native testID/test_id or text target. expected=visible requires a measured viewport intersection; expected=present only requires the mounted tree.",
+      "description": "Wait for a React Native testID/test_id or text target. expected=visible requires a measured viewport intersection; expected=present only requires the mounted tree. expected=absent also passes when every mounted match is on a screen the navigator hides (an inactive tab, or a non-modal stack screen under a non-modal screen) and a native observation does not show the test ID.",
       "schema": {
         "type": "object",
         "properties": {
@@ -960,7 +960,8 @@
             "type": "boolean"
           },
           "timeout_ms": {
-            "type": "number"
+            "type": "number",
+            "description": "Total wait in milliseconds (default 10000). The hidden-screen check for expected=absent needs about 3 s of it left for its native read."
           }
         },
         "additionalProperties": false
@@ -2613,7 +2614,7 @@
       "execution_capabilities": ["app-mutation", "external-mutation"]
     },
     "metamask.perps.close_orders": {
-      "description": "mobile Cancel selected live Perps open orders on the app's current Perps network, with no mainnet opt-in. Use selector/mode params to cancel matching symbols or all orders. Pre-existing testnet state on an account the task owns is converged, not refused.",
+      "description": "mobile Cancel selected live Perps open orders on the app's current Perps network; on mainnet only orders this run placed, anything else is refused before dispatch. Use selector/mode params to cancel matching symbols or all orders. Pre-existing testnet state on an account the task owns is converged, not refused.",
       "schema": {
         "type": "object",
         "properties": {
@@ -2818,11 +2819,24 @@
               "mainnet",
               "testnet"
             ],
-            "description": "Perps network the order must run on; checked against the live app. A mainnet order must declare mainnet."
+            "description": "Perps network the order must run on; checked against the live app. A mainnet order must declare mainnet: a market order with close_immediately, or a dev1 limit order away from the market at 1x and at most $15 notional at its limit price, which the run cancels at its end (success, failure or abort) and records in perps-mainnet-orders.json."
           },
           "close_immediately": {
             "type": "boolean",
-            "description": "Close the opened position again inside this action and prove the market flat (market orders only; the market must start flat). Required on mainnet, with leverage 1 and notional at most $15 from a fixture account."
+            "description": "Close the opened position again inside this action and prove the market flat (market orders only; the market must start flat). A mainnet market order needs it, with leverage 1 and notional at most $15 from a fixture account; a mainnet limit order takes none (the run cancels it at its end)."
+          },
+          "order_type": {
+            "type": "string",
+            "enum": ["market", "limit"],
+            "default": "market"
+          },
+          "limit_price": {
+            "type": ["string", "number"],
+            "description": "Limit price; omit it to offset from the current market price by offset_pct."
+          },
+          "offset_pct": {
+            "type": "number",
+            "description": "Limit price offset from the current market price when limit_price is omitted (default -30 for a buy, 30 for a sell)."
           }
         },
         "additionalProperties": false
@@ -3141,7 +3155,7 @@
       "execution_capabilities": ["app-mutation", "external-mutation"]
     },
     "metamask.perps.ensure_orders": {
-      "description": "mobile Converge selected testnet orders to open or absent, then independently assert the final state. Creating an open state uses a resting limit order, which the shared Perps order rule refuses on mainnet; cancelling acts on the app's current Perps network with no mainnet opt-in. Pre-existing testnet state on an account the task owns is converged, not refused.",
+      "description": "mobile Converge selected testnet orders to open or absent, then independently assert the final state. Creating an open state uses a resting limit order, which the shared Perps order rule refuses on mainnet; cancelling acts on the app's current Perps network, and on mainnet only on orders this run placed. Pre-existing testnet state on an account the task owns is converged, not refused.",
       "schema": {
         "type": "object",
         "properties": {
@@ -3510,7 +3524,7 @@
           "network": {
             "type": "string",
             "enum": ["testnet", "mainnet"],
-            "description": "Requested Perps network. Omit it to keep the current network unless a named profile supplies one."
+            "description": "Requested Perps network; network=mainnet is the opt-in for mainnet orders. Omit it to keep the current network unless a named profile supplies one."
           },
           "market": {
             "type": "string",
@@ -3544,7 +3558,7 @@
           "trading_state": {
             "type": "string",
             "enum": ["preserve", "clean_selected_testnet_market"],
-            "description": "High-level state policy. The clean policy removes matching positions, ordinary orders, TWAPs, and Chase sessions only on testnet and preserves them on mainnet."
+            "description": "High-level state policy that sets positions, orders and strategies itself: preserve leaves them untouched, and the clean policy removes matching positions, ordinary orders, TWAPs, and Chase sessions only on testnet and preserves them on mainnet. Cannot be combined with an explicit positions, orders or strategies request other than false; omit it to apply such a request."
           },
           "positions": {
             "description": "Requested position state. Omit it or pass false to preserve current positions.",
@@ -3768,7 +3782,19 @@
             "additionalProperties": false
           },
           "timeout_ms": {
-            "type": "number"
+            "type": "number",
+            "minimum": 1,
+            "description": "Overall budget for the whole start state setup: every phase runs inside it. Default 30000 on emulators and simulators, 60000 on a physical Android device (explicit non-emulator adb serial), or MM_HARNESS_MOBILE_SETUP_TIMEOUT_MS. A phase that runs out fails as DOMAIN_SETUP_DEADLINE naming the phase."
+          },
+          "navigation_timeout_ms": {
+            "type": "number",
+            "minimum": 1,
+            "description": "Slice of the setup budget for each navigation phase. Default 15000 on emulators and simulators, 30000 on a physical Android device, or MM_HARNESS_MOBILE_NAVIGATION_TIMEOUT_MS."
+          },
+          "cdp_timeout_ms": {
+            "type": "number",
+            "minimum": 1,
+            "description": "Limit for each bridge command inside the setup, capped by the phase deadline. Default 15000. A command that hits it fails as CDP_TIMEOUT (CDP_COMMAND_TIMEOUT in the message)."
           },
           "side": {
             "type": "string",
@@ -4104,7 +4130,19 @@
             "additionalProperties": false
           },
           "timeout_ms": {
-            "type": "number"
+            "type": "number",
+            "minimum": 1,
+            "description": "Overall budget for the whole teardown: every phase runs inside it. Default 30000 on emulators and simulators, 60000 on a physical Android device (explicit non-emulator adb serial), or MM_HARNESS_MOBILE_SETUP_TIMEOUT_MS. A phase that runs out fails as DOMAIN_SETUP_DEADLINE naming the phase."
+          },
+          "navigation_timeout_ms": {
+            "type": "number",
+            "minimum": 1,
+            "description": "Slice of the setup budget for each navigation phase. Default 15000 on emulators and simulators, 30000 on a physical Android device, or MM_HARNESS_MOBILE_NAVIGATION_TIMEOUT_MS."
+          },
+          "cdp_timeout_ms": {
+            "type": "number",
+            "minimum": 1,
+            "description": "Limit for each bridge command inside the setup, capped by the phase deadline. Default 15000. A command that hits it fails as CDP_TIMEOUT (CDP_COMMAND_TIMEOUT in the message)."
           },
           "side": {
             "type": "string",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@deeeed/metamask-harness",
-  "version": "0.86.2",
+  "version": "0.88.0",
   "type": "module",
   "bin": {
     "mm-harness": "bin/mm-harness"
@@ -21,15 +21,15 @@
     "check:syntax": "find . -name '*.mjs' -print0 | xargs -0 -n1 node --check"
   },
   "dependencies": {
-    "@farmslot/adapter-node": "0.1.7",
-    "@farmslot/adapter-rn": "0.19.4",
-    "@farmslot/adapter-sdk": "0.8.1",
-    "@farmslot/adapter-web": "0.6.5",
-    "@farmslot/agent-runtime": "0.17.1",
+    "@farmslot/adapter-node": "0.1.11",
+    "@farmslot/adapter-rn": "0.19.9",
+    "@farmslot/adapter-sdk": "0.8.5",
+    "@farmslot/adapter-web": "0.6.9",
+    "@farmslot/agent-runtime": "0.17.5",
     "@farmslot/handoff": "^0.3.1",
-    "@farmslot/protocol": "0.35.0",
-    "@farmslot/recipe-cli": "0.11.1",
-    "@farmslot/recipe-runner": "0.28.1",
+    "@farmslot/protocol": "0.37.0",
+    "@farmslot/recipe-cli": "0.12.0",
+    "@farmslot/recipe-runner": "0.28.5",
     "agent-device": "0.19.3",
     "commander": "^12.0.0",
     "esbuild": "0.28.1",
@@ -55,6 +55,11 @@
   },
   "main": "./dist/index.js",
   "exports": {
+    "./provider": {
+      "development": "./src/provider.ts",
+      "import": "./dist/provider.js",
+      "default": "./dist/provider.js"
+    },
     ".": {
       "development": "./src/index.ts",
       "import": "./dist/index.js",

package/scripts/validate-release-report.mjs

@@ -1,7 +1,8 @@
 #!/usr/bin/env node
 import fs from 'node:fs';
 import path from 'node:path';
-import { validateRecipeArtifactPackage } from '@farmslot/protocol';
+import { recipeTraceEntries, validateRecipeArtifactPackage } from '@farmslot/protocol';
+import { CAPTURE_EVIDENCE_INCOMPLETE, onlyCaptureInterrupted } from '@farmslot/recipe-runner';
 
 const options = parseArgs(process.argv.slice(2));
 const inventoryPath = path.resolve(options.inventory);
@@ -71,6 +72,7 @@ for (const entry of verdictCases) {
     errors.push(`${entry.id ?? '<missing-id>'} must include evidence`);
     continue;
   }
+  const interruptedRuns = [];
   for (const evidence of entry.evidence) {
     if (typeof evidence !== 'string' || evidence.length === 0) {
       errors.push(`${entry.id ?? '<missing-id>'} contains invalid evidence`);
@@ -121,7 +123,8 @@ for (const entry of verdictCases) {
         }
         documents[required] = requiredDocument.value;
       }
-      if (documents['summary.json'] && entry.status === 'PASS' && documents['summary.json'].status !== 'pass') {
+      const captureInterrupted = onlyCaptureInterrupted(recipeTraceEntries(documents['trace.json']) ?? []);
+      if (documents['summary.json'] && entry.status === 'PASS' && documents['summary.json'].status !== 'pass' && !captureInterrupted) {
         errors.push(`${entry.id ?? '<missing-id>'} PASS evidence summary.json must have status pass: ${resolved}`);
       }
       if (requiredDocuments.every((required) => required in documents)) {
@@ -147,11 +150,20 @@ for (const entry of verdictCases) {
             .join(', ');
           errors.push(`${entry.id ?? '<missing-id>'} evidence directory is not a valid Recipe Protocol artifact package: ${resolved} (${codes})`);
         }
+        interruptedRuns.push(captureInterrupted);
       }
     } else if (!stat.isFile()) {
       errors.push(`${entry.id ?? '<missing-id>'} evidence must be a file or run directory: ${resolved}`);
     }
   }
+  // Retained interrupted attempts must not override a retry's product verdict.
+  if (interruptedRuns.length > 0 && interruptedRuns.every(Boolean)) {
+    if (entry.status !== 'BLOCKED_ENVIRONMENT') {
+      errors.push(`${entry.id ?? '<missing-id>'} evidence failed only because its recording was interrupted (CAPTURE_INTERRUPTED): report BLOCKED_ENVIRONMENT, ${CAPTURE_EVIDENCE_INCOMPLETE}, with the partial video, not ${entry.status}`);
+    } else if (!entry.summary.includes(CAPTURE_EVIDENCE_INCOMPLETE)) {
+      errors.push(`${entry.id ?? '<missing-id>'} summary must say "${CAPTURE_EVIDENCE_INCOMPLETE}" for a case whose recordings were interrupted`);
+    }
+  }
 }
 
 const counts = Object.fromEntries([...allowedStatuses].map((status) => [

package/site/assets/help-recipes.json

@@ -201,7 +201,8 @@
         "Analytics: verify the collection destination and existing consent before capture. Changing participation or marketing consent requires operator approval and restoration of the prior settings; a recipe's hardcoded opt-in is not permission. Missing collection prerequisites are not missing product events.",
         "Flake risk: wait on observable state instead of sleeping, keep device and runtime identity explicit, and never overwrite a prior run's artifacts.",
         "Mutation: require an independent state assertion and identity-bound receipt. Bind activity to that operation's accepted ID; an old row or nonzero count is insufficient. Opening and closing orders have different IDs. Schedule authorized cleanup in teardown, verify its result, and report any residue. Testnet cleanup closes what the task opened and never silently undoes another task's live proof on a reserved account or market.",
-        "Testnet account state: on a testnet account or market the task owns for its run (the slot's fixture account, or one reserved with exclusive_testnet_market=true), converging is an explicit, ownership-scoped setup choice, not a violation. Use the team library's perps.ensure-market-state or perps.clean-market-testnet, or on Mobile and Core metamask.perps.ensure_positions / metamask.perps.ensure_orders (mode=all for strays), record what was closed in the trace, then prove. A pre-existing position or order on an account the task owns is never a reason to block, skip, or refuse, and a fixture that demands an empty account must converge rather than refuse. Extension ensure_* and teardown_state only assert; converge there through an explicitly funded leaf action. Never broaden cleanup onto an account or market the task does not own; obtain an owned one instead of blocking. On mainnet, orders follow the shared Perps order rule (PERPS_MAINNET_ORDER_REFUSED), Core needs CORE_PERPS_ALLOW_MAINNET_WRITES=1 for any write, Mobile cleanup acts on whatever network the app is on, and Extension refuses mainnet writes other than orders and `close_visible_position` in the fixture-policy lane.",
+        "Testnet account state: on a testnet account or market the task owns for its run (the slot's fixture account, or one reserved with exclusive_testnet_market=true), converging is an explicit, ownership-scoped setup choice, not a violation. Use the team library's perps.ensure-market-state or perps.clean-market-testnet, or on Mobile and Core metamask.perps.ensure_positions / metamask.perps.ensure_orders (mode=all for strays), record what was closed in the trace, then prove. A pre-existing position or order on an account the task owns is never a reason to block, skip, or refuse, and a fixture that demands an empty account must converge rather than refuse. Extension ensure_* and teardown_state only assert; converge there through an explicitly funded leaf action. Never broaden cleanup onto an account or market the task does not own; obtain an owned one instead of blocking. On mainnet, orders follow the shared Perps order rule (PERPS_MAINNET_ORDER_REFUSED), Core needs CORE_PERPS_ALLOW_MAINNET_WRITES=1 for any write, Mobile close_positions acts on whatever network the app is on, close_orders there cancels only the run's own orders, and Extension refuses mainnet writes other than orders, `close_visible_position` and cancelling the run's own orders in the fixture-policy lane.",
+        "Rate limits: RATE_LIMITED means Hyperliquid answered 429 Too Many Requests and bounded retries did not get through; it says nothing about the product. Wait a minute and rerun a read. A placement left outcome-unknown is never rerun before the market's positions, orders and fills are checked.",
         {
           "text": "An accepted order can fill before the UI updates. Preserve its receipt even when the expected resting-order assertion fails; one empty read does not prove cleanup. Check the read API's cache and freshness options before adding polling. Reconcile returned margin against the pre-placement balance. A successful close does not turn the original failed proof into a pass.",
           "adapters": [