{"type":"agent-native.clip.context","version":2,"webmcp":{"protocol":"WebMCP","scope":"page-local","tools":[{"name":"clips-get-context","title":"Get clip context","description":"Read clip metadata, readiness, transcript status, and the HTTP API URLs. This works without a browser.","inputSchema":{"type":"object","properties":{},"additionalProperties":false},"annotations":{"readOnlyHint":true,"untrustedContentHint":true}},{"name":"clips-get-transcript","title":"Get clip transcript","description":"Read bounded timestamped transcript segments. Use nextStartIndex to page through long transcripts without losing overlapping segments. For complete transcript text, follow sourceUrl to the HTTP transcript endpoint.","inputSchema":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this start time."},"endMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this end time."},"startIndex":{"type":"integer","minimum":0,"description":"Stable segment index for pagination; prefer nextStartIndex from the previous page."},"maxSegments":{"type":"integer","minimum":1,"maximum":50,"description":"Maximum number of transcript segments to return."}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"untrustedContentHint":true}},{"name":"clips-get-frame","title":"Get clip frame","description":"Get the clip's JPEG frame URL at a requested timestamp. Fetch the returned imageUrl as an image to inspect the video; its scoped access query is already included. On an error, inspect the JSON failureKind and nextStep: access needs a Share with agents link, media means the stored recording media could not be retrieved, processing means the frame request failed and may be retried or reported, unsupported means this source cannot provide frames through Clips and must not be retried, and expired means the owner must update the clip expiry before creating another link.","inputSchema":{"type":"object","properties":{"atMs":{"type":"number","minimum":0,"description":"Video timestamp in milliseconds; fetch the returned imageUrl as an image to inspect the frame."}},"required":["atMs"],"additionalProperties":false},"availability":"ready","annotations":{"readOnlyHint":true,"untrustedContentHint":true}},{"name":"clips-get-frames","title":"Get clip frame sheet","description":"Get a URL for one JPEG contact sheet of up to 36 consecutive video frames, to watch motion such as an animation or a transition in a single image. Cells read left to right, then top to bottom; cell i shows the frame on screen at startMs + i * 1000 / fps. Solid black cells at the end of the grid hold no frame. The range can span at most 60 seconds. On an error, inspect the JSON failureKind and nextStep; failureKind=request means the range or fps was out of bounds, and nextStep names a request that fits.","inputSchema":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Time of the first frame, in milliseconds."},"endMs":{"type":"number","minimum":0,"description":"End of the range in milliseconds (exclusive). Omit it to get up to 36 frames from startMs, stopping at 60 seconds or the end of the clip."},"fps":{"type":"number","minimum":0.1,"maximum":60,"description":"Frames per second to sample. Omit it to use the video's own frame rate, so every frame in the range appears once."}},"required":["startMs"],"additionalProperties":false},"availability":"ready","annotations":{"readOnlyHint":true,"untrustedContentHint":true}}],"instructions":"For browser-independent access from any HTTP client, fetch the exact agentContextUrl and use its apis.context, apis.transcript, apis.frame, and apis.frames URLs. Keep the clip id and any agent_access query parameter exactly as provided on every request. Read the complete transcript from apis.transcript and fetch recommendedFrames[].url (or apis.frame.urlTemplate with atMs) as images to inspect the video. To watch motion, fetch apis.frames.urlTemplate: one contact-sheet image of consecutive frames. Do not stop at the transcript when frames are available. For any non-2xx response, inspect the JSON failureKind, error, and nextStep. If a context or transcript request has failureKind=access, explain that this link does not grant access; for a private clip, ask its owner to open the Clips Share menu, choose Share with agents, and send the generated link. If a frame request has failureKind=media while the context or transcript still works, explain that the stored recording media could not be retrieved and another share link will not fix it. If a frame request has failureKind=processing, follow nextStep; extraction or temporary storage failures do not mean the clip is missing or inaccessible. If a request has failureKind=expired, ask the owner to extend or remove the clip's expiry in the Share menu, save it, then create and send a new Share with agents link. If this clip page is already open in a WebMCP-capable browser, list its page tools for bounded read-only access; clips-get-transcript may omit fullText or return a truncated result, and its sourceUrl points to the HTTP transcript. Use nextStartIndex for transcript pagination so overlapping segments are not lost.","http":{"schema_version":"v1","browserRequired":false,"tools":[{"name":"clips-get-context","title":"Get clip context","description":"Read clip metadata, readiness, transcript status, and the HTTP API URLs. This works without a browser.","inputSchema":{"type":"object","properties":{},"additionalProperties":false},"annotations":{"readOnlyHint":true,"untrustedContentHint":true},"parameters":{"type":"object","properties":{},"additionalProperties":false},"endpoint":"https://clips.agent-native.com/api/agent-context.json?id=77mChMteEQ6M","method":"GET","responseType":"application/json"},{"name":"clips-get-transcript","title":"Get clip transcript","description":"Read bounded timestamped transcript segments. Use nextStartIndex to page through long transcripts without losing overlapping segments. For complete transcript text, follow sourceUrl to the HTTP transcript endpoint.","inputSchema":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this start time."},"endMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this end time."},"startIndex":{"type":"integer","minimum":0,"description":"Stable segment index for pagination; prefer nextStartIndex from the previous page."},"maxSegments":{"type":"integer","minimum":1,"maximum":50,"description":"Maximum number of transcript segments to return."}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"untrustedContentHint":true},"parameters":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this start time."},"endMs":{"type":"number","minimum":0,"description":"Only return segments that overlap this end time."},"startIndex":{"type":"integer","minimum":0,"description":"Stable segment index for pagination; prefer nextStartIndex from the previous page."},"maxSegments":{"type":"integer","minimum":1,"maximum":50,"description":"Maximum number of transcript segments to return."}},"additionalProperties":false},"endpoint":"https://clips.agent-native.com/api/agent-transcript.json?id=77mChMteEQ6M","method":"GET","responseType":"application/json"},{"name":"clips-get-frame","title":"Get clip frame","description":"Get the clip's JPEG frame URL at a requested timestamp. Fetch the returned imageUrl as an image to inspect the video; its scoped access query is already included. On an error, inspect the JSON failureKind and nextStep: access needs a Share with agents link, media means the stored recording media could not be retrieved, processing means the frame request failed and may be retried or reported, unsupported means this source cannot provide frames through Clips and must not be retried, and expired means the owner must update the clip expiry before creating another link.","inputSchema":{"type":"object","properties":{"atMs":{"type":"number","minimum":0,"description":"Video timestamp in milliseconds; fetch the returned imageUrl as an image to inspect the frame."}},"required":["atMs"],"additionalProperties":false},"availability":"ready","annotations":{"readOnlyHint":true,"untrustedContentHint":true},"parameters":{"type":"object","properties":{"atMs":{"type":"number","minimum":0,"description":"Video timestamp in milliseconds; fetch the returned imageUrl as an image to inspect the frame."}},"required":["atMs"],"additionalProperties":false},"endpoint":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs={atMs}","method":"GET","responseType":"image/jpeg"},{"name":"clips-get-frames","title":"Get clip frame sheet","description":"Get a URL for one JPEG contact sheet of up to 36 consecutive video frames, to watch motion such as an animation or a transition in a single image. Cells read left to right, then top to bottom; cell i shows the frame on screen at startMs + i * 1000 / fps. Solid black cells at the end of the grid hold no frame. The range can span at most 60 seconds. On an error, inspect the JSON failureKind and nextStep; failureKind=request means the range or fps was out of bounds, and nextStep names a request that fits.","inputSchema":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Time of the first frame, in milliseconds."},"endMs":{"type":"number","minimum":0,"description":"End of the range in milliseconds (exclusive). Omit it to get up to 36 frames from startMs, stopping at 60 seconds or the end of the clip."},"fps":{"type":"number","minimum":0.1,"maximum":60,"description":"Frames per second to sample. Omit it to use the video's own frame rate, so every frame in the range appears once."}},"required":["startMs"],"additionalProperties":false},"availability":"ready","annotations":{"readOnlyHint":true,"untrustedContentHint":true},"parameters":{"type":"object","properties":{"startMs":{"type":"number","minimum":0,"description":"Time of the first frame, in milliseconds."},"endMs":{"type":"number","minimum":0,"description":"End of the range in milliseconds (exclusive). Omit it to get up to 36 frames from startMs, stopping at 60 seconds or the end of the clip."},"fps":{"type":"number","minimum":0.1,"maximum":60,"description":"Frames per second to sample. Omit it to use the video's own frame rate, so every frame in the range appears once."}},"required":["startMs"],"additionalProperties":false},"endpoint":"https://clips.agent-native.com/api/agent-frames.jpg?id=77mChMteEQ6M&startMs={startMs}&endMs={endMs}&fps={fps}","method":"GET","responseType":"image/jpeg"}]}},"instructions":["Use transcript.segments for timestamped spoken context.","Use the HTTP URLs in apis for browser-independent access. For complete transcript text, use apis.transcript. If this clip page is already open in a WebMCP-capable browser, list its read-only page tools for bounded access; WebMCP transcript results may omit fullText or be truncated, so follow sourceUrl for the complete transcript.","Keep the exact id and any agent_access query parameter from each supplied URL. When a frame request returns JSON, read failureKind, error, and nextStep. If failureKind=access, explain that the link does not grant access; for a private clip, ask the owner to open the Clips Share menu, choose Share with agents, and send the generated link. If failureKind=media, explain that the stored media could not be retrieved and another share link will not fix it. If failureKind=processing, follow nextStep. When clip.agentReadiness.state is preparing, wait clip.agentReadiness.retryAfterSeconds and fetch apis.context.url before retrying; do not treat the frame as missing media. If failureKind=expired, ask the owner to extend or remove the clip's expiry in the Share menu, save it, then create and send a new Share with agents link. Report the failing frame URL without its agent_access value.","This clip is readable as both text (transcript) and images (JPEG frames) — you can hear AND see it.","To SEE the screen, GET apis.frame.urlTemplate with atMs (returns image/jpeg). Start with recommendedFrames, then fetch additional frames around transcript timestamps that matter for the task. Pass a comment's videoTimestampMs as atMs to see the frame its author was looking at. Older comments were saved rounded down to the whole second, so their frame can be up to a second early. To watch motion (an animation, a transition, a cursor path), GET apis.frames.urlTemplate with startMs, endMs and fps: one JPEG contact sheet of up to 36 consecutive frames, read left to right, then top to bottom. Cell i shows the frame on screen at startMs + i * 1000 / fps. Leave out fps to use the video's own frame rate, so every frame appears once; leave out endMs for up to 36 frames from startMs. Solid black cells at the end of the grid hold no frame. If it returns failureKind=request, follow nextStep."],"clip":{"id":"77mChMteEQ6M","title":"apprentice-film","description":"Apprentice learns issue-triage workflows from Maya’s typed actions, turns demonstrations into editable playbooks, and tests them against past issues before acting. It starts in shadow mode, requires approvals before promotion, and keeps actions undoable; incorrect decisions can be overridden, with corrections added as regression tests. Apprentice imports GitHub issues read-only and exports playbooks as skill files for other agents.","publicPageUrl":"https://clips.agent-native.com/share/77mChMteEQ6M","sourceProvider":null,"thumbnailUrl":"https://cdn.builder.io/api/v1/image/assets%2F75c7560bf5de4eaca0e0e03da6bc7485%2Fa976800bad144777ab2123a25804710e","animatedThumbnailUrl":null,"durationMs":180800,"duration":"181s","width":1920,"height":1080,"hasAudio":true,"hasCamera":false,"status":"ready","agentReadiness":{"state":"ready","retryAfterSeconds":null,"instruction":null},"createdAt":"2026-11-09T20:18:27.743Z","updatedAt":"2026-11-09T20:22:59.010Z"},"apis":{"mcp":{"tool":"get-recording-frames","connectUrl":"https://clips.agent-native.com/mcp/connect"},"context":{"method":"GET","url":"https://clips.agent-native.com/api/agent-context.json?id=77mChMteEQ6M"},"transcript":{"method":"GET","url":"https://clips.agent-native.com/api/agent-transcript.json?id=77mChMteEQ6M"},"frame":{"method":"GET","urlTemplate":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs={timestampMs}","responseType":"image/jpeg","query":{"atMs":"Video timestamp in milliseconds. The endpoint returns image/jpeg."}},"frames":{"method":"GET","urlTemplate":"https://clips.agent-native.com/api/agent-frames.jpg?id=77mChMteEQ6M&startMs={startMs}&endMs={endMs}&fps={fps}","responseType":"image/jpeg","maxFrames":36,"maxSpanMs":60000,"query":{"startMs":"Time of the first frame, in milliseconds.","endMs":"Optional end of the range in milliseconds (exclusive). Without it, the sheet holds up to 36 frames from startMs, stopping at 60 seconds or the end of the clip.","fps":"Optional frames per second, from 0.1 to 60. Without it, the sheet uses the video's own frame rate."}}},"transcript":{"status":"ready","language":"en","failureReason":null,"retryAfterSeconds":null,"fullText":"Maya has clicked this exact sequence a hundred times. Today, it'll be 101. This is Apprentice. Maya maintains an open source project. Every morning brings a fresh pile of issues. Half of them just say, it's broken, no version, no steps, so she does the same thing every single time. Label it, ask for a repro, assign it, next. An agent could help if she wrote it a prompt and trusted it. Apprentice learns the job by watching her do it. Then it has to prove it before it's trusted to act. These are real open issues from the Agent Native repo on GitHub. Apprentice imports them read-only. It never writes back. Maya triages issue 31 the way she always does, needs info, bug, request a repro, assigning to triage. Then 32, every click is a typed action, so each one is recorded as a demonstration. On 33, it's already predicting her next step. She just presses tab. And it noticed the pattern. It even sees that her teammate Bob handles these differently and lets the team choose. One click turns it into a playbook, she can read and edit. Before she trusts it, it replays her history, 29 issues, 17 matched, 14 exactly how she would do them. She loosens one condition and the false positives jump from 3 to 7. So she keeps it tight. Then it runs in shadow mode. Nothing changes until she approves. Five approvals in a row earn a promotion, draft then act. At act, new issues are handled on their own. Four handled, each one with an undo. Who taught you this? It points straight to her own demonstrations on issues 31 and 32. Now a real one from GitHub. Issue 8918 is a feature request, not a bug. The rule got it wrong, so Meyer overrides it. That correction becomes a regression test. In the playbook steps back to draft until it passes. Here's why this works. In Agent Native, a button click and an agent call the exact same typed action. So every click is already a clean demonstration, no screen scraping. And the trust ladder is the framework's own approval gate. Back to Maya's morning. Apprentice handles the repeats and shows her only what needs her. And every playbook exports as a skill file that any agent can use. Maya closes her laptop before lunch. Apprentice, show it once, delegate forever.","segments":[{"startMs":0,"endMs":2333,"timestamp":"0:00","range":"0:00-0:02","text":"Maya has clicked this exact sequence a"},{"startMs":2333,"endMs":3000,"timestamp":"0:02","range":"0:02-0:03","text":"hundred times."},{"startMs":3000,"endMs":6000,"timestamp":"0:03","range":"0:03-0:06","text":"Today, it'll be 101."},{"startMs":8000,"endMs":10000,"timestamp":"0:08","range":"0:08-0:10","text":"This is Apprentice."},{"startMs":13000,"endMs":17000,"timestamp":"0:13","range":"0:13-0:17","text":"Maya maintains an open source project."},{"startMs":17000,"endMs":21375,"timestamp":"0:17","range":"0:17-0:21","text":"Every morning brings a fresh pile of"},{"startMs":21375,"endMs":22000,"timestamp":"0:21","range":"0:21-0:22","text":"issues."},{"startMs":22000,"endMs":23750,"timestamp":"0:22","range":"0:22-0:23","text":"Half of them just say,"},{"startMs":23750,"endMs":25150,"timestamp":"0:23","range":"0:23-0:25","text":"it's broken, no version,"},{"startMs":25150,"endMs":27600,"timestamp":"0:25","range":"0:25-0:27","text":"no steps, so she does the same"},{"startMs":27600,"endMs":29000,"timestamp":"0:27","range":"0:27-0:29","text":"thing every single time."},{"startMs":29000,"endMs":31667,"timestamp":"0:29","range":"0:29-0:31","text":"Label it, ask for a repro,"},{"startMs":31667,"endMs":33000,"timestamp":"0:31","range":"0:31-0:33","text":"assign it, next."},{"startMs":33000,"endMs":35154,"timestamp":"0:33","range":"0:33-0:35","text":"An agent could help if she wrote"},{"startMs":35154,"endMs":37000,"timestamp":"0:35","range":"0:35-0:37","text":"it a prompt and trusted it."},{"startMs":40000,"endMs":42333,"timestamp":"0:40","range":"0:40-0:42","text":"Apprentice learns the job by watching her"},{"startMs":42333,"endMs":43000,"timestamp":"0:42","range":"0:42-0:43","text":"do it."},{"startMs":43000,"endMs":44909,"timestamp":"0:43","range":"0:43-0:44","text":"Then it has to prove it before"},{"startMs":44909,"endMs":46000,"timestamp":"0:44","range":"0:44-0:46","text":"it's trusted to act."},{"startMs":48000,"endMs":50333,"timestamp":"0:48","range":"0:48-0:50","text":"These are real open issues from the"},{"startMs":50333,"endMs":52000,"timestamp":"0:50","range":"0:50-0:52","text":"Agent Native repo on GitHub."},{"startMs":52000,"endMs":54000,"timestamp":"0:52","range":"0:52-0:54","text":"Apprentice imports them read-only."},{"startMs":54000,"endMs":56000,"timestamp":"0:54","range":"0:54-0:56","text":"It never writes back."},{"startMs":56000,"endMs":58722,"timestamp":"0:56","range":"0:56-0:58","text":"Maya triages issue 31 the way she"},{"startMs":58722,"endMs":60278,"timestamp":"0:58","range":"0:58-1:00","text":"always does, needs info,"},{"startMs":60278,"endMs":61834,"timestamp":"1:00","range":"1:00-1:01","text":"bug, request a repro,"},{"startMs":61834,"endMs":63000,"timestamp":"1:01","range":"1:01-1:03","text":"assigning to triage."},{"startMs":63000,"endMs":65625,"timestamp":"1:03","range":"1:03-1:05","text":"Then 32, every click is a typed"},{"startMs":65625,"endMs":68250,"timestamp":"1:05","range":"1:05-1:08","text":"action, so each one is recorded as"},{"startMs":68250,"endMs":69000,"timestamp":"1:08","range":"1:08-1:09","text":"a demonstration."},{"startMs":69000,"endMs":71625,"timestamp":"1:09","range":"1:09-1:11","text":"On 33, it's already predicting her next"},{"startMs":71625,"endMs":72000,"timestamp":"1:11","range":"1:11-1:12","text":"step."},{"startMs":72000,"endMs":73000,"timestamp":"1:12","range":"1:12-1:13","text":"She just presses tab."},{"startMs":75000,"endMs":77000,"timestamp":"1:15","range":"1:15-1:17","text":"And it noticed the pattern."},{"startMs":77000,"endMs":79800,"timestamp":"1:17","range":"1:17-1:19","text":"It even sees that her teammate Bob"},{"startMs":79800,"endMs":82600,"timestamp":"1:19","range":"1:19-1:22","text":"handles these differently and lets the team"},{"startMs":82600,"endMs":83000,"timestamp":"1:22","range":"1:22-1:23","text":"choose."},{"startMs":83000,"endMs":85333,"timestamp":"1:23","range":"1:23-1:25","text":"One click turns it into a playbook,"},{"startMs":85333,"endMs":87000,"timestamp":"1:25","range":"1:25-1:27","text":"she can read and edit."},{"startMs":87000,"endMs":88263,"timestamp":"1:27","range":"1:27-1:28","text":"Before she trusts it,"},{"startMs":88263,"endMs":89526,"timestamp":"1:28","range":"1:28-1:29","text":"it replays her history,"},{"startMs":89526,"endMs":90789,"timestamp":"1:29","range":"1:29-1:30","text":"29 issues, 17 matched,"},{"startMs":90789,"endMs":93000,"timestamp":"1:30","range":"1:30-1:33","text":"14 exactly how she would do them."},{"startMs":93000,"endMs":95692,"timestamp":"1:33","range":"1:33-1:35","text":"She loosens one condition and the false"},{"startMs":95692,"endMs":98000,"timestamp":"1:35","range":"1:35-1:38","text":"positives jump from 3 to 7."},{"startMs":98000,"endMs":99000,"timestamp":"1:38","range":"1:38-1:39","text":"So she keeps it tight."},{"startMs":101000,"endMs":103000,"timestamp":"1:41","range":"1:41-1:43","text":"Then it runs in shadow mode."},{"startMs":103000,"endMs":105000,"timestamp":"1:43","range":"1:43-1:45","text":"Nothing changes until she approves."},{"startMs":105000,"endMs":107545,"timestamp":"1:45","range":"1:45-1:47","text":"Five approvals in a row earn a"},{"startMs":107545,"endMs":109000,"timestamp":"1:47","range":"1:47-1:49","text":"promotion, draft then act."},{"startMs":110000,"endMs":112333,"timestamp":"1:50","range":"1:50-1:52","text":"At act, new issues are handled on"},{"startMs":112333,"endMs":113000,"timestamp":"1:52","range":"1:52-1:53","text":"their own."},{"startMs":113000,"endMs":116000,"timestamp":"1:53","range":"1:53-1:56","text":"Four handled, each one with an undo."},{"startMs":116000,"endMs":118000,"timestamp":"1:56","range":"1:56-1:58","text":"Who taught you this?"},{"startMs":118000,"endMs":120917,"timestamp":"1:58","range":"1:58-2:00","text":"It points straight to her own demonstrations"},{"startMs":120917,"endMs":123000,"timestamp":"2:00","range":"2:00-2:03","text":"on issues 31 and 32."},{"startMs":123000,"endMs":125000,"timestamp":"2:03","range":"2:03-2:05","text":"Now a real one from GitHub."},{"startMs":125000,"endMs":127667,"timestamp":"2:05","range":"2:05-2:07","text":"Issue 8918 is a feature request,"},{"startMs":127667,"endMs":129000,"timestamp":"2:07","range":"2:07-2:09","text":"not a bug."},{"startMs":129000,"endMs":130667,"timestamp":"2:09","range":"2:09-2:10","text":"The rule got it wrong,"},{"startMs":130667,"endMs":132000,"timestamp":"2:10","range":"2:10-2:12","text":"so Meyer overrides it."},{"startMs":132000,"endMs":135000,"timestamp":"2:12","range":"2:12-2:15","text":"That correction becomes a regression test."},{"startMs":135000,"endMs":137800,"timestamp":"2:15","range":"2:15-2:17","text":"In the playbook steps back to draft"},{"startMs":137800,"endMs":139000,"timestamp":"2:17","range":"2:17-2:19","text":"until it passes."},{"startMs":140000,"endMs":142000,"timestamp":"2:20","range":"2:20-2:22","text":"Here's why this works."},{"startMs":142000,"endMs":143000,"timestamp":"2:22","range":"2:22-2:23","text":"In Agent Native,"},{"startMs":143000,"endMs":145333,"timestamp":"2:23","range":"2:23-2:25","text":"a button click and an agent call"},{"startMs":145333,"endMs":147000,"timestamp":"2:25","range":"2:25-2:27","text":"the exact same typed action."},{"startMs":147000,"endMs":149545,"timestamp":"2:27","range":"2:27-2:29","text":"So every click is already a clean"},{"startMs":149545,"endMs":151000,"timestamp":"2:29","range":"2:29-2:31","text":"demonstration, no screen scraping."},{"startMs":151000,"endMs":153800,"timestamp":"2:31","range":"2:31-2:33","text":"And the trust ladder is the framework's"},{"startMs":153800,"endMs":155000,"timestamp":"2:33","range":"2:33-2:35","text":"own approval gate."},{"startMs":158000,"endMs":160000,"timestamp":"2:38","range":"2:38-2:40","text":"Back to Maya's morning."},{"startMs":160000,"endMs":163182,"timestamp":"2:40","range":"2:40-2:43","text":"Apprentice handles the repeats and shows her"},{"startMs":163182,"endMs":165000,"timestamp":"2:43","range":"2:43-2:45","text":"only what needs her."},{"startMs":165000,"endMs":167692,"timestamp":"2:45","range":"2:45-2:47","text":"And every playbook exports as a skill"},{"startMs":167692,"endMs":170000,"timestamp":"2:47","range":"2:47-2:50","text":"file that any agent can use."},{"startMs":175000,"endMs":178000,"timestamp":"2:55","range":"2:55-2:58","text":"Maya closes her laptop before lunch."},{"startMs":178000,"endMs":179910,"timestamp":"2:58","range":"2:58-2:59","text":"Apprentice, show it once,"},{"startMs":179910,"endMs":180865,"timestamp":"2:59","range":"2:59-3:00","text":"delegate forever."}],"segmentCount":80},"comments":[],"commentCount":0,"commentsTruncated":false,"reactions":[],"reactionCount":0,"reactionsTruncated":false,"chapters":[],"recommendedFrames":[{"atMs":350,"timestamp":"0:00","reason":"opening frame","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=350"},{"atMs":8000,"timestamp":"0:08","reason":"transcript: This is Apprentice.","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=8000"},{"atMs":13000,"timestamp":"0:13","reason":"transcript: Maya maintains an open source project.","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=13000"},{"atMs":21375,"timestamp":"0:21","reason":"transcript: issues.","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=21375"},{"atMs":27600,"timestamp":"0:27","reason":"transcript: thing every single time.","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=27600"},{"atMs":33000,"timestamp":"0:33","reason":"transcript: An agent could help if she wrote","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=33000"},{"atMs":40000,"timestamp":"0:40","reason":"transcript: Apprentice learns the job by watching her","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=40000"},{"atMs":45200,"timestamp":"0:45","reason":"25% mark","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=45200"},{"atMs":90400,"timestamp":"1:30","reason":"50% mark","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=90400"},{"atMs":135600,"timestamp":"2:15","reason":"75% mark","url":"https://clips.agent-native.com/api/agent-frame.jpg?id=77mChMteEQ6M&atMs=135600"}],"bugReport":null,"browserDiagnostics":null,"appActivity":null,"ctas":[]}