@aspectlylabs/mikka 0.0.128 → 0.0.130

package/cli.js

@@ -4457,7 +4457,7 @@ import { join as join4 } from "node:path";
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -4517,7 +4517,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts
@@ -4981,12 +4981,15 @@ var captureCompleteResponse = z23.object({ capture: captureView });
 var screenCaptureRequest = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), title }).strict(),
   z23.object({ action: z23.literal("start_recording"), title, maxSeconds: z23.number().int().min(5).max(RECORDING_MAX_SECONDS).optional() }).strict(),
-  z23.object({ action: z23.literal("stop_recording") }).strict()
+  z23.object({ action: z23.literal("stop_recording") }).strict(),
+  // A step on the recording's timeline: an action's own description, or a moment the agent marks.
+  z23.object({ action: z23.literal("step"), label: title, source: z23.enum(["action", "mark"]) }).strict()
 ]);
 var screenCaptureResponse = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), capture: captureView }),
   z23.object({ action: z23.literal("start_recording"), recordingId: z23.uuid(), stopsAt: z23.iso.datetime() }),
-  z23.object({ action: z23.literal("stop_recording"), capture: captureView })
+  z23.object({ action: z23.literal("stop_recording"), capture: captureView }),
+  z23.object({ action: z23.literal("step"), recorded: z23.boolean() })
 ]);
 var capturePaths = {
   begin: (runId) => `/v1/runs/${encodeURIComponent(runId)}/captures`,
@@ -14052,7 +14055,12 @@ async function detectComputerDriver(binary = "cua-driver") {
     return null;
   }
 }
-var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording" };
+var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording", mark_recording_step: "step" };
+var cleanStep = (value) => typeof value === "string" ? value.replace(/\s+/g, " ").trim() : "";
+function actionStep(value) {
+  const label2 = cleanStep(value?.arguments?.step);
+  return label2 && label2.length <= 200 ? label2 : null;
+}
 function normalizeCaptureCall(value) {
   if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k))) throw new Error("Unknown call fields");
@@ -14062,6 +14070,10 @@ function normalizeCaptureCall(value) {
   const keys = Object.keys(args).sort().join(",");
   const title2 = typeof args.title === "string" ? args.title.replace(/\s+/g, " ").trim() : "";
   if (action === "stop_recording" && !keys) return { action };
+  if (action === "step" && keys === "label") {
+    const label2 = cleanStep(args.label);
+    if (label2 && label2.length <= 200) return { action, label: label2, source: "mark" };
+  }
   if (action === "screenshot" && keys === "title" && title2 && title2.length <= 200) return { action, title: title2 };
   if (action === "start_recording" && ["title", "maxSeconds,title"].includes(keys) && title2 && title2.length <= 200 && (args.maxSeconds === void 0 || Number.isInteger(args.maxSeconds) && args.maxSeconds >= 5 && args.maxSeconds <= 300))
     return { action, title: title2, ...args.maxSeconds === void 0 ? {} : { maxSeconds: args.maxSeconds } };
@@ -14072,9 +14084,12 @@ function normalizeCall(value) {
     throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k)))
     throw new Error("Unknown call fields");
-  const { name, arguments: args = {} } = value;
-  if (!args || typeof args !== "object" || Array.isArray(args))
+  const { name, arguments: given = {} } = value;
+  if (!given || typeof given !== "object" || Array.isArray(given))
     throw new Error("Invalid arguments");
+  const { step: described, ...args } = given;
+  if (Object.hasOwn(given, "step") && (name === "get_desktop_state" || !cleanStep(described) || cleanStep(described).length > 200))
+    throw new Error("Unsupported tool or arguments");
   const keys = Object.keys(args).sort().join(",");
   if (name === "get_desktop_state" && !keys) return { name, arguments: {} };
   if (name === "click" && keys === "x,y" && [args.x, args.y].every((v2) => Number.isInteger(v2) && v2 >= 0 && v2 <= 1e5))
@@ -14201,6 +14216,8 @@ async function executeDriver(driver, input, display, leaseGuard, assertLive = ()
 // src/computer/broker.mjs
 var ComputerBroker = class {
   grants = /* @__PURE__ */ new Map();
+  // runId → when its screen recording stops by itself; only set while one runs.
+  recordingUntil = /* @__PURE__ */ new Map();
   pending = /* @__PURE__ */ new Map();
   sockets = /* @__PURE__ */ new Set();
   suspended = false;
@@ -14279,6 +14296,24 @@ var ComputerBroker = class {
     if (!grant || signal?.aborted || !this.permitted(grant.runId))
       return Promise.reject(new Error("Not permitted"));
     const call = normalizeCall(value);
+    const step = actionStep(value);
+    return this.requestCall(token, grant, call, signal).then((result) => {
+      this.markStep(grant.runId, step);
+      return result;
+    });
+  }
+  /** While this run records, a done action's description becomes a step on the recording (best effort; never fails the action). */
+  markStep(runId, label2) {
+    const until = this.recordingUntil.get(runId);
+    if (!label2 || !this.capture || until === void 0) return;
+    if (this.now() > until) {
+      this.recordingUntil.delete(runId);
+      return;
+    }
+    void Promise.resolve().then(() => this.capture(runId, { action: "step", label: label2, source: "action" })).catch(() => {
+    });
+  }
+  requestCall(token, grant, call, signal) {
     const actionId = randomUUID9();
     const assertLive = () => {
       if (this.suspended || signal?.aborted || this.grants.get(token) !== grant || this.driver?.fingerprint !== grant.fingerprint || !this.permitted(grant.runId))
@@ -14331,7 +14366,10 @@ var ComputerBroker = class {
       return { error: "Not permitted" };
     const call = normalizeCaptureCall(value);
     try {
-      return { result: await this.capture(grant.runId, call, signal) };
+      const result = await this.capture(grant.runId, call, signal);
+      if (call.action === "start_recording" && typeof result?.stopsAt === "string") this.recordingUntil.set(grant.runId, Date.parse(result.stopsAt));
+      if (call.action === "stop_recording") this.recordingUntil.delete(grant.runId);
+      return { result };
     } catch (error) {
       const detail = typeof error?.safeMessage === "string" ? error.safeMessage.slice(0, 300) : void 0;
       return { error: "Capture failed", ...detail ? { detail } : {} };

package/client.js

@@ -3719,7 +3719,7 @@ import { join } from "node:path";
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -3779,7 +3779,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts

package/computer/driver.mjs

@@ -28,6 +28,13 @@ export async function detectComputerDriver(binary = "cua-driver") {
     return null;
   }
 }
+// What an action does, in the person's words. While a screen recording runs it
+// becomes a step on the recording's timeline (Figma 13), so it is never
+// coordinates, keys or a tool name.
+const step = {
+  type: "string", minLength: 1, maxLength: 200,
+  description: "What this action does, in plain words for the person watching (e.g. \"Click Sign in\", \"点击登录按钮\"). Shown on the screen recording's timeline; never coordinates or tool names.",
+};
 export const TOOLS = [
   {
     name: "get_desktop_state",
@@ -47,8 +54,9 @@ export const TOOLS = [
       properties: {
         x: { type: "integer", minimum: 0, maximum: 100000 },
         y: { type: "integer", minimum: 0, maximum: 100000 },
+        step,
       },
-      required: ["x", "y"],
+      required: ["x", "y", "step"],
       additionalProperties: false,
     },
   },
@@ -58,8 +66,8 @@ export const TOOLS = [
       "Type exactly the requested text into the focused desktop control, subject to the computer permission policy.",
     inputSchema: {
       type: "object",
-      properties: { text: { type: "string", minLength: 1, maxLength: 8000 } },
-      required: ["text"],
+      properties: { text: { type: "string", minLength: 1, maxLength: 8000 }, step },
+      required: ["text", "step"],
       additionalProperties: false,
     },
   },
@@ -68,8 +76,8 @@ export const TOOLS = [
     description: "Press the requested key and modifiers on the focused desktop, subject to the computer permission policy.",
     inputSchema: {
       type: "object",
-      properties: { key: { type: "string" }, modifiers: { type: "array", items: { type: "string", enum: ["ctrl", "shift", "alt", "super"] }, maxItems: 4, uniqueItems: true } },
-      required: ["key"], additionalProperties: false,
+      properties: { key: { type: "string" }, modifiers: { type: "array", items: { type: "string", enum: ["ctrl", "shift", "alt", "super"] }, maxItems: 4, uniqueItems: true }, step },
+      required: ["key", "step"], additionalProperties: false,
     },
   },
   {
@@ -77,8 +85,8 @@ export const TOOLS = [
     description: "Scroll the requested amount at a desktop coordinate, subject to the computer permission policy.",
     inputSchema: {
       type: "object",
-      properties: { direction: { type: "string", enum: ["up", "down", "left", "right"] }, amount: { type: "integer", minimum: 1, maximum: 50 }, x: { type: "integer", minimum: 0, maximum: 100000 }, y: { type: "integer", minimum: 0, maximum: 100000 } },
-      required: ["direction", "amount", "x", "y"], additionalProperties: false,
+      properties: { direction: { type: "string", enum: ["up", "down", "left", "right"] }, amount: { type: "integer", minimum: 1, maximum: 50 }, x: { type: "integer", minimum: 0, maximum: 100000 }, y: { type: "integer", minimum: 0, maximum: 100000 }, step },
+      required: ["direction", "amount", "x", "y", "step"], additionalProperties: false,
     },
   },
 ];
@@ -108,8 +116,20 @@ export const CAPTURE_TOOLS = [
     description: "Stop the screen recording and keep it with the thread.",
     inputSchema: { type: "object", properties: {}, additionalProperties: false },
   },
+  {
+    name: "mark_recording_step",
+    description:
+      "While recording, mark a moment that is not one of your actions (e.g. \"The first screen appears\", \"首屏出现\") on the recording's timeline. Each action's `step` is marked by itself.",
+    inputSchema: { type: "object", properties: { label: step }, required: ["label"], additionalProperties: false },
+  },
 ];
-const CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording" };
+const CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording", mark_recording_step: "step" };
+const cleanStep = (value) => (typeof value === "string" ? value.replace(/\s+/g, " ").trim() : "");
+/** The action's own description for the recording timeline, or null. */
+export function actionStep(value) {
+  const label = cleanStep(value?.arguments?.step);
+  return label && label.length <= 200 ? label : null;
+}
 export function normalizeCaptureCall(value) {
   if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k))) throw new Error("Unknown call fields");
@@ -119,6 +139,10 @@ export function normalizeCaptureCall(value) {
   const keys = Object.keys(args).sort().join(",");
   const title = typeof args.title === "string" ? args.title.replace(/\s+/g, " ").trim() : "";
   if (action === "stop_recording" && !keys) return { action };
+  if (action === "step" && keys === "label") {
+    const label = cleanStep(args.label);
+    if (label && label.length <= 200) return { action, label, source: "mark" };
+  }
   if (action === "screenshot" && keys === "title" && title && title.length <= 200) return { action, title };
   if (action === "start_recording" && ["title", "maxSeconds,title"].includes(keys) && title && title.length <= 200 && (args.maxSeconds === undefined || (Number.isInteger(args.maxSeconds) && args.maxSeconds >= 5 && args.maxSeconds <= 300)))
     return { action, title, ...(args.maxSeconds === undefined ? {} : { maxSeconds: args.maxSeconds }) };
@@ -130,9 +154,14 @@ export function normalizeCall(value) {
     throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k)))
     throw new Error("Unknown call fields");
-  const { name, arguments: args = {} } = value;
-  if (!args || typeof args !== "object" || Array.isArray(args))
+  const { name, arguments: given = {} } = value;
+  if (!given || typeof given !== "object" || Array.isArray(given))
     throw new Error("Invalid arguments");
+  // `step` is the person's description, kept by the broker for the recording
+  // timeline; it never reaches the driver or a review.
+  const { step: described, ...args } = given;
+  if (Object.hasOwn(given, "step") && (name === "get_desktop_state" || !cleanStep(described) || cleanStep(described).length > 200))
+    throw new Error("Unsupported tool or arguments");
   const keys = Object.keys(args).sort().join(",");
   if (name === "get_desktop_state" && !keys) return { name, arguments: {} };
   if (

package/http-link.js

@@ -2727,12 +2727,15 @@ var captureCompleteResponse = z9.object({ capture: captureView });
 var screenCaptureRequest = z9.discriminatedUnion("action", [
   z9.object({ action: z9.literal("screenshot"), title }).strict(),
   z9.object({ action: z9.literal("start_recording"), title, maxSeconds: z9.number().int().min(5).max(RECORDING_MAX_SECONDS).optional() }).strict(),
-  z9.object({ action: z9.literal("stop_recording") }).strict()
+  z9.object({ action: z9.literal("stop_recording") }).strict(),
+  // A step on the recording's timeline: an action's own description, or a moment the agent marks.
+  z9.object({ action: z9.literal("step"), label: title, source: z9.enum(["action", "mark"]) }).strict()
 ]);
 var screenCaptureResponse = z9.discriminatedUnion("action", [
   z9.object({ action: z9.literal("screenshot"), capture: captureView }),
   z9.object({ action: z9.literal("start_recording"), recordingId: z9.uuid(), stopsAt: z9.iso.datetime() }),
-  z9.object({ action: z9.literal("stop_recording"), capture: captureView })
+  z9.object({ action: z9.literal("stop_recording"), capture: captureView }),
+  z9.object({ action: z9.literal("step"), recorded: z9.boolean() })
 ]);
 var capturePaths = {
   begin: (runId) => `/v1/runs/${encodeURIComponent(runId)}/captures`,
@@ -4118,7 +4121,7 @@ var InvalidIdError = class extends DaemonError {
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -4178,7 +4181,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts

package/login.js

@@ -3726,7 +3726,7 @@ var ServerUnreachableError = class extends DaemonError {
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -3786,7 +3786,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts
@@ -3844,12 +3844,15 @@ var captureCompleteResponse = z23.object({ capture: captureView });
 var screenCaptureRequest = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), title }).strict(),
   z23.object({ action: z23.literal("start_recording"), title, maxSeconds: z23.number().int().min(5).max(RECORDING_MAX_SECONDS).optional() }).strict(),
-  z23.object({ action: z23.literal("stop_recording") }).strict()
+  z23.object({ action: z23.literal("stop_recording") }).strict(),
+  // A step on the recording's timeline: an action's own description, or a moment the agent marks.
+  z23.object({ action: z23.literal("step"), label: title, source: z23.enum(["action", "mark"]) }).strict()
 ]);
 var screenCaptureResponse = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), capture: captureView }),
   z23.object({ action: z23.literal("start_recording"), recordingId: z23.uuid(), stopsAt: z23.iso.datetime() }),
-  z23.object({ action: z23.literal("stop_recording"), capture: captureView })
+  z23.object({ action: z23.literal("stop_recording"), capture: captureView }),
+  z23.object({ action: z23.literal("step"), recorded: z23.boolean() })
 ]);
 
 // src/run-mailbox.ts

package/main.js

@@ -3773,7 +3773,7 @@ import { join as join5 } from "node:path";
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -3833,7 +3833,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts
@@ -4120,12 +4120,15 @@ var captureCompleteResponse = z7.object({ capture: captureView });
 var screenCaptureRequest = z7.discriminatedUnion("action", [
   z7.object({ action: z7.literal("screenshot"), title }).strict(),
   z7.object({ action: z7.literal("start_recording"), title, maxSeconds: z7.number().int().min(5).max(RECORDING_MAX_SECONDS).optional() }).strict(),
-  z7.object({ action: z7.literal("stop_recording") }).strict()
+  z7.object({ action: z7.literal("stop_recording") }).strict(),
+  // A step on the recording's timeline: an action's own description, or a moment the agent marks.
+  z7.object({ action: z7.literal("step"), label: title, source: z7.enum(["action", "mark"]) }).strict()
 ]);
 var screenCaptureResponse = z7.discriminatedUnion("action", [
   z7.object({ action: z7.literal("screenshot"), capture: captureView }),
   z7.object({ action: z7.literal("start_recording"), recordingId: z7.uuid(), stopsAt: z7.iso.datetime() }),
-  z7.object({ action: z7.literal("stop_recording"), capture: captureView })
+  z7.object({ action: z7.literal("stop_recording"), capture: captureView }),
+  z7.object({ action: z7.literal("step"), recorded: z7.boolean() })
 ]);
 
 // ../protocol/src/accounts.ts
@@ -13127,7 +13130,12 @@ async function detectComputerDriver(binary = "cua-driver") {
     return null;
   }
 }
-var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording" };
+var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording", mark_recording_step: "step" };
+var cleanStep = (value) => typeof value === "string" ? value.replace(/\s+/g, " ").trim() : "";
+function actionStep(value) {
+  const label2 = cleanStep(value?.arguments?.step);
+  return label2 && label2.length <= 200 ? label2 : null;
+}
 function normalizeCaptureCall(value) {
   if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k))) throw new Error("Unknown call fields");
@@ -13137,6 +13145,10 @@ function normalizeCaptureCall(value) {
   const keys = Object.keys(args).sort().join(",");
   const title2 = typeof args.title === "string" ? args.title.replace(/\s+/g, " ").trim() : "";
   if (action === "stop_recording" && !keys) return { action };
+  if (action === "step" && keys === "label") {
+    const label2 = cleanStep(args.label);
+    if (label2 && label2.length <= 200) return { action, label: label2, source: "mark" };
+  }
   if (action === "screenshot" && keys === "title" && title2 && title2.length <= 200) return { action, title: title2 };
   if (action === "start_recording" && ["title", "maxSeconds,title"].includes(keys) && title2 && title2.length <= 200 && (args.maxSeconds === void 0 || Number.isInteger(args.maxSeconds) && args.maxSeconds >= 5 && args.maxSeconds <= 300))
     return { action, title: title2, ...args.maxSeconds === void 0 ? {} : { maxSeconds: args.maxSeconds } };
@@ -13147,9 +13159,12 @@ function normalizeCall(value) {
     throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k)))
     throw new Error("Unknown call fields");
-  const { name, arguments: args = {} } = value;
-  if (!args || typeof args !== "object" || Array.isArray(args))
+  const { name, arguments: given = {} } = value;
+  if (!given || typeof given !== "object" || Array.isArray(given))
     throw new Error("Invalid arguments");
+  const { step: described, ...args } = given;
+  if (Object.hasOwn(given, "step") && (name === "get_desktop_state" || !cleanStep(described) || cleanStep(described).length > 200))
+    throw new Error("Unsupported tool or arguments");
   const keys = Object.keys(args).sort().join(",");
   if (name === "get_desktop_state" && !keys) return { name, arguments: {} };
   if (name === "click" && keys === "x,y" && [args.x, args.y].every((v2) => Number.isInteger(v2) && v2 >= 0 && v2 <= 1e5))
@@ -13276,6 +13291,8 @@ async function executeDriver(driver, input, display, leaseGuard, assertLive = ()
 // src/computer/broker.mjs
 var ComputerBroker = class {
   grants = /* @__PURE__ */ new Map();
+  // runId → when its screen recording stops by itself; only set while one runs.
+  recordingUntil = /* @__PURE__ */ new Map();
   pending = /* @__PURE__ */ new Map();
   sockets = /* @__PURE__ */ new Set();
   suspended = false;
@@ -13354,6 +13371,24 @@ var ComputerBroker = class {
     if (!grant || signal?.aborted || !this.permitted(grant.runId))
       return Promise.reject(new Error("Not permitted"));
     const call = normalizeCall(value);
+    const step = actionStep(value);
+    return this.requestCall(token, grant, call, signal).then((result) => {
+      this.markStep(grant.runId, step);
+      return result;
+    });
+  }
+  /** While this run records, a done action's description becomes a step on the recording (best effort; never fails the action). */
+  markStep(runId, label2) {
+    const until = this.recordingUntil.get(runId);
+    if (!label2 || !this.capture || until === void 0) return;
+    if (this.now() > until) {
+      this.recordingUntil.delete(runId);
+      return;
+    }
+    void Promise.resolve().then(() => this.capture(runId, { action: "step", label: label2, source: "action" })).catch(() => {
+    });
+  }
+  requestCall(token, grant, call, signal) {
     const actionId = randomUUID8();
     const assertLive = () => {
       if (this.suspended || signal?.aborted || this.grants.get(token) !== grant || this.driver?.fingerprint !== grant.fingerprint || !this.permitted(grant.runId))
@@ -13406,7 +13441,10 @@ var ComputerBroker = class {
       return { error: "Not permitted" };
     const call = normalizeCaptureCall(value);
     try {
-      return { result: await this.capture(grant.runId, call, signal) };
+      const result = await this.capture(grant.runId, call, signal);
+      if (call.action === "start_recording" && typeof result?.stopsAt === "string") this.recordingUntil.set(grant.runId, Date.parse(result.stopsAt));
+      if (call.action === "stop_recording") this.recordingUntil.delete(grant.runId);
+      return { result };
     } catch (error) {
       const detail = typeof error?.safeMessage === "string" ? error.safeMessage.slice(0, 300) : void 0;
       return { error: "Capture failed", ...detail ? { detail } : {} };

package/mikka-release-manifest.json

@@ -1,11 +1,11 @@
 {
   "version": 1,
-  "sourceSha": "f4a767de25bc5f9854e4df728bd43253b7486d2a",
+  "sourceSha": "3436cdaa0961b829bba4be118a6277081cd8f3b9",
   "repository": "alexj11324/Mikka",
-  "buildRunId": "41271217664",
+  "buildRunId": "41271504440",
   "buildKind": "github-actions-main",
   "sourceLockSha256": "c99f1e4f641d8e004cfe807653478e10f91e53fcf47102add5f4c154cdc1cb57",
-  "createdAt": "2026-11-01T04:09:41.956Z",
+  "createdAt": "2026-11-01T04:17:35.833Z",
   "sbom": "sbom.cdx.json",
   "files": [
     {
@@ -40,18 +40,18 @@
     },
     {
       "path": "cli.js",
-      "size": 632263,
-      "sha256": "a77e0b676915675d9c124a15f3475431b011bc9ee4b4f8bcd8b0ff14cd771aac"
+      "size": 634339,
+      "sha256": "8080156886d23b77b66fdbe44a5f900c75bf92f34907ce655e4f811284b25c08"
     },
     {
       "path": "client.js",
       "size": 139865,
-      "sha256": "32227b674d07e94bab3fb4aa3477969b9614856fde60153249c1ae5a9b58cf8c"
+      "sha256": "4d1de4cea6f074a4c084c56b41e111b6c10670d651353e83243c4a6afe85444b"
     },
     {
       "path": "computer/driver.mjs",
-      "size": 12788,
-      "sha256": "ce40e50779ef36e3f5f333b1315e7c53c4746f5ac60c0ffa4aae16a03bcf39ad"
+      "size": 14571,
+      "sha256": "7368022aad90b87836a9bcebe726a9e683ff12fdc6a9ab4d98fd2aa641a3910d"
     },
     {
       "path": "computer/proxy.mjs",
@@ -170,8 +170,8 @@
     },
     {
       "path": "http-link.js",
-      "size": 180835,
-      "sha256": "d90dbdc702f4319c057c815402bc3ed0bef88d0f85a5318d61d2ac49876922a7"
+      "size": 181108,
+      "sha256": "b7d56b8c5aaddbc53194935eb437c235538c0a396230af038e4b7bfadddf7b94"
     },
     {
       "path": "log.js",
@@ -180,13 +180,13 @@
     },
     {
       "path": "login.js",
-      "size": 144321,
-      "sha256": "0179f53fd594ec6bb08adb82b9de7d910842e714d51c5113c1364c5590d1ac3c"
+      "size": 144600,
+      "sha256": "202ddf2106bd7e48d8161b7b085cf78bfe35fa284b6fc881f41fa693f2192f4b"
     },
     {
       "path": "main.js",
-      "size": 577894,
-      "sha256": "ebc871ec42a6efb044ea7b9826e13466af8da730c831a1938b806106b1e894b7"
+      "size": 579964,
+      "sha256": "6071ee3133ed667bf9174552f2ebeda443235386a619cbb9507097e28de4a190"
     },
     {
       "path": "mikka-mcp.mjs",
@@ -195,23 +195,23 @@
     },
     {
       "path": "mikka.js",
-      "size": 632279,
-      "sha256": "07b6e0bd34daa6962160a529f3a248dc0b94d68b3e715553041a2b4bcf6d68a4"
+      "size": 634355,
+      "sha256": "0e314fe228d3c4b41d6fa36a0060c845978f776db53bed7a00fc72c3a1e78c5b"
     },
     {
       "path": "npm-shrinkwrap.json",
       "size": 78117,
-      "sha256": "94a855dabf6c3fb403b7b9b0bd72517ec64caab9f6a487bf18e815fc44f21f3b"
+      "sha256": "fdefc0f2623430cca795db0fd5998fa79675e09f3a035ffa1036b88b5808e77e"
     },
     {
       "path": "package.json",
       "size": 1247,
-      "sha256": "de5493ee58154e642ef6fdd6f142bd2faad8dc5eaabd28811876575ec3103204"
+      "sha256": "06ac83bdc11be2c08997e4b51f5154c7e0cf2fa02d26e98ee30dd711bc0d2f90"
     },
     {
       "path": "paths.js",
       "size": 4138,
-      "sha256": "ec4475513e9a3c8e8a2bed522744fbc80f4053a5bb8f1f7723ed7d4dfc79a83a"
+      "sha256": "72f9c8961b1731a6be12d22fc07c050245eed9ff5a329fe1c41ed71e99119227"
     },
     {
       "path": "sandbox.js",
@@ -221,7 +221,7 @@
     {
       "path": "sbom.cdx.json",
       "size": 144314,
-      "sha256": "b774018d22cb5a3c386663faede01f08a68bb4e47fe26e33ffbedb221d2afac3"
+      "sha256": "bc4d2fa8b01ddf8726be293d47cc23e9fc5c5c94fc25732bab8beb574c68d78c"
     }
   ]
 }

package/mikka.js

@@ -4458,7 +4458,7 @@ import { join as join4 } from "node:path";
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -4518,7 +4518,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts
@@ -4982,12 +4982,15 @@ var captureCompleteResponse = z23.object({ capture: captureView });
 var screenCaptureRequest = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), title }).strict(),
   z23.object({ action: z23.literal("start_recording"), title, maxSeconds: z23.number().int().min(5).max(RECORDING_MAX_SECONDS).optional() }).strict(),
-  z23.object({ action: z23.literal("stop_recording") }).strict()
+  z23.object({ action: z23.literal("stop_recording") }).strict(),
+  // A step on the recording's timeline: an action's own description, or a moment the agent marks.
+  z23.object({ action: z23.literal("step"), label: title, source: z23.enum(["action", "mark"]) }).strict()
 ]);
 var screenCaptureResponse = z23.discriminatedUnion("action", [
   z23.object({ action: z23.literal("screenshot"), capture: captureView }),
   z23.object({ action: z23.literal("start_recording"), recordingId: z23.uuid(), stopsAt: z23.iso.datetime() }),
-  z23.object({ action: z23.literal("stop_recording"), capture: captureView })
+  z23.object({ action: z23.literal("stop_recording"), capture: captureView }),
+  z23.object({ action: z23.literal("step"), recorded: z23.boolean() })
 ]);
 var capturePaths = {
   begin: (runId) => `/v1/runs/${encodeURIComponent(runId)}/captures`,
@@ -14053,7 +14056,12 @@ async function detectComputerDriver(binary = "cua-driver") {
     return null;
   }
 }
-var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording" };
+var CAPTURE_ACTIONS = { capture_screenshot: "screenshot", start_screen_recording: "start_recording", stop_screen_recording: "stop_recording", mark_recording_step: "step" };
+var cleanStep = (value) => typeof value === "string" ? value.replace(/\s+/g, " ").trim() : "";
+function actionStep(value) {
+  const label2 = cleanStep(value?.arguments?.step);
+  return label2 && label2.length <= 200 ? label2 : null;
+}
 function normalizeCaptureCall(value) {
   if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k))) throw new Error("Unknown call fields");
@@ -14063,6 +14071,10 @@ function normalizeCaptureCall(value) {
   const keys = Object.keys(args).sort().join(",");
   const title2 = typeof args.title === "string" ? args.title.replace(/\s+/g, " ").trim() : "";
   if (action === "stop_recording" && !keys) return { action };
+  if (action === "step" && keys === "label") {
+    const label2 = cleanStep(args.label);
+    if (label2 && label2.length <= 200) return { action, label: label2, source: "mark" };
+  }
   if (action === "screenshot" && keys === "title" && title2 && title2.length <= 200) return { action, title: title2 };
   if (action === "start_recording" && ["title", "maxSeconds,title"].includes(keys) && title2 && title2.length <= 200 && (args.maxSeconds === void 0 || Number.isInteger(args.maxSeconds) && args.maxSeconds >= 5 && args.maxSeconds <= 300))
     return { action, title: title2, ...args.maxSeconds === void 0 ? {} : { maxSeconds: args.maxSeconds } };
@@ -14073,9 +14085,12 @@ function normalizeCall(value) {
     throw new Error("Invalid tool call");
   if (Object.keys(value).some((k) => !["name", "arguments"].includes(k)))
     throw new Error("Unknown call fields");
-  const { name, arguments: args = {} } = value;
-  if (!args || typeof args !== "object" || Array.isArray(args))
+  const { name, arguments: given = {} } = value;
+  if (!given || typeof given !== "object" || Array.isArray(given))
     throw new Error("Invalid arguments");
+  const { step: described, ...args } = given;
+  if (Object.hasOwn(given, "step") && (name === "get_desktop_state" || !cleanStep(described) || cleanStep(described).length > 200))
+    throw new Error("Unsupported tool or arguments");
   const keys = Object.keys(args).sort().join(",");
   if (name === "get_desktop_state" && !keys) return { name, arguments: {} };
   if (name === "click" && keys === "x,y" && [args.x, args.y].every((v2) => Number.isInteger(v2) && v2 >= 0 && v2 <= 1e5))
@@ -14202,6 +14217,8 @@ async function executeDriver(driver, input, display, leaseGuard, assertLive = ()
 // src/computer/broker.mjs
 var ComputerBroker = class {
   grants = /* @__PURE__ */ new Map();
+  // runId → when its screen recording stops by itself; only set while one runs.
+  recordingUntil = /* @__PURE__ */ new Map();
   pending = /* @__PURE__ */ new Map();
   sockets = /* @__PURE__ */ new Set();
   suspended = false;
@@ -14280,6 +14297,24 @@ var ComputerBroker = class {
     if (!grant || signal?.aborted || !this.permitted(grant.runId))
       return Promise.reject(new Error("Not permitted"));
     const call = normalizeCall(value);
+    const step = actionStep(value);
+    return this.requestCall(token, grant, call, signal).then((result) => {
+      this.markStep(grant.runId, step);
+      return result;
+    });
+  }
+  /** While this run records, a done action's description becomes a step on the recording (best effort; never fails the action). */
+  markStep(runId, label2) {
+    const until = this.recordingUntil.get(runId);
+    if (!label2 || !this.capture || until === void 0) return;
+    if (this.now() > until) {
+      this.recordingUntil.delete(runId);
+      return;
+    }
+    void Promise.resolve().then(() => this.capture(runId, { action: "step", label: label2, source: "action" })).catch(() => {
+    });
+  }
+  requestCall(token, grant, call, signal) {
     const actionId = randomUUID9();
     const assertLive = () => {
       if (this.suspended || signal?.aborted || this.grants.get(token) !== grant || this.driver?.fingerprint !== grant.fingerprint || !this.permitted(grant.runId))
@@ -14332,7 +14367,10 @@ var ComputerBroker = class {
       return { error: "Not permitted" };
     const call = normalizeCaptureCall(value);
     try {
-      return { result: await this.capture(grant.runId, call, signal) };
+      const result = await this.capture(grant.runId, call, signal);
+      if (call.action === "start_recording" && typeof result?.stopsAt === "string") this.recordingUntil.set(grant.runId, Date.parse(result.stopsAt));
+      if (call.action === "stop_recording") this.recordingUntil.delete(grant.runId);
+      return { result };
     } catch (error) {
       const detail = typeof error?.safeMessage === "string" ? error.safeMessage.slice(0, 300) : void 0;
       return { error: "Capture failed", ...detail ? { detail } : {} };

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "@aspectlylabs/mikka",
-  "version": "0.0.128",
+  "version": "0.0.130",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@aspectlylabs/mikka",
-      "version": "0.0.128",
+      "version": "0.0.130",
       "license": "AGPL-3.0-only",
       "dependencies": {
         "@agentclientprotocol/sdk": "1.9.0",

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "@aspectlylabs/mikka",
-  "version": "0.0.128",
+  "version": "0.0.130",
   "description": "Local Mikka daemon: runs agents through the user's own CLIs",
-  "gitHead": "f4a767de25bc5f9854e4df728bd43253b7486d2a",
+  "gitHead": "3436cdaa0961b829bba4be118a6277081cd8f3b9",
   "type": "module",
   "private": false,
   "license": "AGPL-3.0-only",

package/paths.js

@@ -6,7 +6,7 @@ import { join } from "node:path";
 // package.json
 var package_default = {
   name: "@mikka/daemon",
-  version: "0.0.128",
+  version: "0.0.130",
   description: "Local Mikka daemon: runs agents through the user's own CLIs",
   type: "module",
   private: true,
@@ -66,7 +66,7 @@ var package_default = {
   engines: {
     node: ">=24"
   },
-  gitHead: "f4a767de25bc5f9854e4df728bd43253b7486d2a"
+  gitHead: "3436cdaa0961b829bba4be118a6277081cd8f3b9"
 };
 
 // src/paths.ts

package/sbom.cdx.json

@@ -2,10 +2,10 @@
   "$schema": "http://cyclonedx.org/schema/bom-1.5.schema.json",
   "bomFormat": "CycloneDX",
   "specVersion": "1.5",
-  "serialNumber": "urn:uuid:627c2ed9-ab34-4cd3-9cc5-5bf71cb7e63f",
+  "serialNumber": "urn:uuid:93681fb3-f996-4fde-9eb6-a30300e52a72",
   "version": 1,
   "metadata": {
-    "timestamp": "2026-11-01T04:09:41.030Z",
+    "timestamp": "2026-11-01T04:17:34.918Z",
     "lifecycles": [
       {
         "phase": "pre-build"
@@ -19,13 +19,13 @@
       }
     ],
     "component": {
-      "bom-ref": "@aspectlylabs/mikka@0.0.128",
+      "bom-ref": "@aspectlylabs/mikka@0.0.130",
       "type": "library",
-      "name": ".retained-build-YJhfqC",
-      "version": "0.0.128",
+      "name": ".retained-build-F1hFpJ",
+      "version": "0.0.130",
       "scope": "required",
       "description": "Local Mikka daemon: runs agents through the user's own CLIs",
-      "purl": "pkg:npm/%40aspectlylabs/mikka@0.0.128",
+      "purl": "pkg:npm/%40aspectlylabs/mikka@0.0.130",
       "properties": [],
       "externalReferences": [
         {
@@ -4627,7 +4627,7 @@
   ],
   "dependencies": [
     {
-      "ref": "@aspectlylabs/mikka@0.0.128",
+      "ref": "@aspectlylabs/mikka@0.0.130",
       "dependsOn": [
         "@agentclientprotocol/sdk@1.9.0",
         "@anthropic-ai/claude-agent-sdk@0.3.305",