yt-briefing 1.2.5 → 1.3.1

package/README.md

@@ -16,8 +16,8 @@ It also gets better the more you use it. You give each summary a quick rating, w
 or not, and from that it learns what to keep showing you and what to drop. Over time the queue
 becomes yours: less noise, more of what you care about.
 
-yt-briefing runs inside [Claude Code](https://claude.com/claude-code). `/yt` opens the briefing in
-a pane next to your chat, and the filtering and summaries run on your own Claude Code login. There
+yt-briefing runs inside [Claude Code](https://claude.com/claude-code). `/yt` runs the briefing in
+your chat, and the filtering and summaries run on your own Claude Code login. There
 is no separate model, provider or LLM key to set up.
 
 ## First run vs later
@@ -120,32 +120,25 @@ it to go deeper, lower it for a quicker pass:
 
 ## Run it
 
-Open your project in Claude Code and type `/yt`. The briefing opens in a pane: the summary, and
-under it four keys.
-
-The keys work while the pane has the focus (it takes it when it opens, `/yt` from an empty
-prompt), and Esc closes it.
-
-| Key | What it does |
-|-----|--------------|
-| `1` OK | Neutral. The video is marked as seen, the next one loads. |
-| `2` Weak | Worthless. The title goes to the channel's skip examples, so the filter learns to drop titles like it. |
-| `3` Research | Ends the loop and hands this video to Claude, see below. |
-| `4` Stop | Closes the pane. The next `/yt` resumes where you stopped. |
-
-The **Comment** field takes anything else. Type what you think in your own words ("too many panel
-shows, skip those") and press Enter: Claude turns it into a standing rule for that channel and
-infers the rating. `? your question` starts research with that question, `stop` closes the pane.
-Claude Code on a phone has no text fields, so there the pane shows the four keys only.
-Where no pane can be seen (`/yt` sent from claude.ai/code or the phone app over Remote Control, or
-a surface that places no panes), `/yt` hands the same loop to Claude in the chat: each summary is
-Claude's message, a short dialog asks for the rating (OK / Weak / Research / Stop), and a typed
-**Other** answer works as the Comment field. This path uses the model, so it counts toward plan usage.
-
-Each step is the engine, not a chat turn: rating a video takes no turn and no context of your
-session, and the next summary is usually ready before you have finished reading the current one.
-The summaries themselves still count toward your Claude plan's usage, like any other Claude Code
-work.
+Open your project in Claude Code and type `/yt`. Each summary arrives as Claude's message in the
+chat, and a short dialog under it asks for the rating:
+
+| Answer | What it does |
+|--------|--------------|
+| OK | Neutral. The video is marked as seen, the next one loads. |
+| Weak | Worthless. The title goes to the channel's skip examples, so the filter learns to drop titles like it. |
+| Research | Ends the loop and hands this video to Claude, see below. |
+| Stop | Ends the loop. The next `/yt` resumes where you stopped. |
+
+The dialog's **Other** field takes anything else. Type what you think in your own words ("too many
+panel shows, skip those"): Claude turns it into a standing rule for that channel and infers the
+rating. `? your question` starts research with that question, `stop` ends the loop. A prompt you
+type in the chat also ends it.
+
+It works the same in the terminal, the desktop app, claude.ai/code and the phone app. Each summary
+is a turn of its own that ends with the summary, so the phone app shows it whole instead of folding
+it into a one-line digest. Claude pastes the summary the engine wrote, so a turn is short, but it
+still counts toward your Claude plan's usage, like the summaries themselves.
 
 `/yt` is a Claude Code mod (a plugin in `.claude/skills/yt-briefing/`). Claude Code loads it on
 its own once you trust the project folder. If `/yt` is not listed, start a fresh session. To
@@ -156,8 +149,8 @@ installs `/yt`, `/yt-transcribe` and `/yt-search`).
 
 Tech channels announce something new every week, and the usual fate is "looks interesting" →
 to-do list → never. So next to OK/Weak there is a third key: **Research**. Press it, or type
-`? your question` into the comment field, and the loop ends there: the pane closes and the video
-lands in your chat with its briefing and the command for its full transcript. Claude works your
+`? your question` into the dialog's Other field, and the loop ends there: the video lands in your
+chat with its briefing and the command for its full transcript. Claude works your
 question with you, against your own codebase if you ask "would this fit my project", against the
 web if the claims need checking. A quick feedback loop instead of a shelf. The video is marked as
 seen, and the next `/yt` resumes the queue right where you broke off.
@@ -166,8 +159,8 @@ seen, and the next `/yt` resumes the queue right where you broke off.
 
 1.0 runs on Claude Code only. The OpenAI-compatible provider and its three `YT_BRIEFING_LLM_*`
 keys are gone (delete them from `.env`), and the chat-driven `/yt` skill with its rating popup is
-replaced by the pane. Run `npx yt-briefing install-skill` once in your project: it installs the
-pane, and removes the old `/yt` skill and the summary-gate hook from `.claude/settings.json`. Your
+replaced by the `/yt` mod. Run `npx yt-briefing install-skill` once in your project: it installs
+the mod, and removes the old `/yt` skill and the summary-gate hook from `.claude/settings.json`. Your
 channels, profiles and ratings in `.yt-briefing/data/` carry over unchanged.
 
 ## Why Claude Code, and nothing else
@@ -179,13 +172,12 @@ is set in your environment, the engine removes it for that call, because Claude
 otherwise bill it as API usage instead of using your login.
 
 The engine still works ahead in the background. It expands channels in parallel and summarizes the
-next video while you rate the current one, so each step is usually ready with no wait. The pane
-only shows what the engine produced and sends your key presses back to it.
+next video while you rate the current one, so each step is usually ready with no wait. The mod
+asks for the rating and records it itself, so the model never has to remember the protocol.
 
-Supporting every agent that reads `SKILL.md` meant a chat loop for the rating: the model pasted
-each summary, asked the question and recorded the answer, every video a full turn, plus a hook to
-make sure the summary was really shown. A Claude Code pane does the same with no model in the loop,
-which is why 1.0 drops the other agents.
+Supporting every agent that reads `SKILL.md` meant leaving the rating to the model: it asked the
+question and recorded the answer, plus a hook to make sure the summary was really shown. A Claude
+Code mod drives the loop itself, which is why 1.0 drops the other agents.
 
 ## Why one transcript at a time
 

package/dist/bootstrap.js

@@ -8,7 +8,7 @@
  *   1. Output language for summaries + ratings   → DATA_DIR/config.json
  *   2. The channels you follow — just a flat list of handles
  *      → DATA_DIR/channels.md, DATA_DIR/state.md, DATA_DIR/channels/<slug>.md
- *   3. Installs /yt (the rating pane) + /yt-transcribe + /yt-search into this Claude Code project
+ *   3. Installs /yt (the rating loop) + /yt-transcribe + /yt-search into this Claude Code project
  *
  * It does NOT touch keys: the one key (YouTube Data API) lives in your project root .env (see README →
  * Setup); the engine reads it at run time. Filters and summaries run on your Claude Code login. Re-running is safe: it warns before overwriting existing data and bails.
@@ -124,7 +124,7 @@ function main() {
         console.log(`\n  ⚠ ${noClaude}`);
     console.log('\n  Next:');
     console.log('    1. Open this folder in Claude Code (trust it when asked).');
-    console.log('    2. Type  /yt  — the briefing opens in a pane.  (/yt-transcribe <url> for one video)\n');
+    console.log('    2. Type  /yt  — the briefing runs in the chat.  (/yt-transcribe <url> for one video)\n');
 }
 try {
     main();

package/dist/cli.js

@@ -6,7 +6,7 @@
  * or Bun), passing remaining args through.
  *
  *   yt-briefing init                       interactive onboarding wizard
- *   yt-briefing install-skill              install /yt (pane) + skills into a Claude Code project
+ *   yt-briefing install-skill              install /yt + skills into a Claude Code project
  *   yt-briefing add|remove <@handle|url>   add or remove channels (also list)
  *   yt-briefing list                       list the channels you follow
  *   yt-briefing sweep [--reset]            advance one step; prints a JSON status line

package/dist/install-skill.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
 /**
- * install-skill — install yt-briefing into a Claude Code project: the `/yt` mod (the rating pane)
+ * install-skill — install yt-briefing into a Claude Code project: the `/yt` mod (the rating loop)
  * plus the `/yt-transcribe` and `/yt-search` skills, all under `<project>/.claude/skills/`.
  *
  *   yt-briefing install-skill        # interactive: which project folder
@@ -13,7 +13,7 @@
 import { installAll, isPackageDevCwd } from "./lib/skill-install.js";
 import { question } from "./lib/prompt.js";
 const ask = (q, def = '') => question(def ? `${q} [${def}]:` : `${q}:`).trim() || def;
-console.log('\n  Install /yt (pane) + /yt-transcribe + /yt-search into a Claude Code project.\n');
+console.log('\n  Install /yt + /yt-transcribe + /yt-search into a Claude Code project.\n');
 console.log('    1) This project (current folder) — recommended');
 console.log('    2) Another project folder\n');
 const other = ask('  Where', '1') === '2';

package/dist/lib/distill.js

@@ -2,8 +2,8 @@
  * Turn a raw rating comment ("too much politics, skip these panels") into what the profile
  * stores: one clean, generalizable rule for `## Notes`, plus the rating it implies.
  *
- * This used to be the agent's job in the chat loop. With the rating loop in a pane there is no
- * agent turn per video, so the engine does it with the same `claude -p` call it uses everywhere.
+ * This used to be the agent's job in the chat loop. The mod records ratings itself, with no agent
+ * step for it, so the engine does it with the same `claude -p` call it uses everywhere.
  */
 import { chat } from "./llm.js";
 export function distillPrompt(comment, video) {

package/dist/lib/skill-install.js

@@ -6,7 +6,7 @@
  * `<project>/.claude/skills/`:
  *
  *   yt-briefing/   the Claude Code mod (a plugin of function hooks) behind `/yt`: the rating loop
- *                  as a pane. Claude Code loads a plugin from the project's skills folder by itself
+ *                  in the chat. Claude Code loads a plugin from the project's skills folder by itself
  *                  (as `yt-briefing@skills-dir`) once the workspace is trusted.
  *   yt-transcribe/ one-shot skill: a single video's transcript → summary.
  *   yt-search/     skill: search within one channel → triage → comparison.
@@ -79,7 +79,7 @@ export function skillBody(name, dist = false) {
         .replace(/data\//g, toProjectRel(DATA_DIR) + '/');
 }
 /**
- * The mod's `hooks/engine.ts`: how the pane runs the engine. `dist=false` is the shipped dev form;
+ * The mod's `hooks/engine.ts`: how the mod runs the engine. `dist=false` is the shipped dev form;
  * `dist=true` points at the compiled scripts, relative to the project root (the session's cwd).
  */
 export function engineModule(dist = false) {
@@ -88,7 +88,7 @@ export function engineModule(dist = false) {
         return shipped;
     const rel = toProjectRel(DIST_DIR);
     return [
-        '// How the pane runs the engine, relative to the project root (the session\'s working directory).',
+        '// How the mod runs the engine, relative to the project root (the session\'s working directory).',
         '// Written by `yt-briefing install-skill`; re-run it after moving the project or switching runtime.',
         `export const engine = (name: string): string[] => ['${runtimeName()}', \`${rel}/\${name}.js\`]`,
         '',

package/dist/yt-sweep.js

@@ -296,6 +296,7 @@ Steps:
    - 2-5 numbered thematic sections × 2-5 sentences each
    - At most 5-8 short quotes from the transcript
    - No timestamps
+   - No section of your own on the material's credibility, sourcing or bias (no "reliability notes", no "caveats"): report what the video says. Only a channel directive can ask for such an assessment.
 3. Apply the channel directives above, if any — they shape the briefing's content, emphasis, and tone, but not the header/subtitle/section skeleton. Put any directive-driven assessment in its own short, clearly-labeled section.
 4. Language: natural ${LANG}. Avoid calques/anglicisms; use foreign words only for proper nouns or established technical terms. Section headers should be verb phrases, not noun stacks.
 5. Output: ONLY the briefing itself OR 'OFFTOPIC: ...'. No preamble, and no meta-commentary about these instructions.`;

package/docs/sync-across-machines.md

@@ -99,7 +99,7 @@ version, so every machine gets the setting):
 }
 ```
 
-The engine runs `after_rate` after every recorded rating, from the pane and from the CLI
+The engine runs `after_rate` after every recorded rating, from `/yt` and from the CLI
 (`yt-briefing rate`) alike. It runs detached from the project root, so a slow push never holds the
 next video, and a failing script never fails the rating (the rating is already on disk). The
 engine itself still never runs git: the command is yours.

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "yt-briefing",
-  "version": "1.2.5",
-  "description": "A self-learning YouTube briefing for Claude Code: it sweeps the channels you follow, filters noise in two stages (title, then transcript), gives you a short briefing of the rest in your language in a pane, and adapts to your ratings — one video at a time.",
+  "version": "1.3.1",
+  "description": "A self-learning YouTube briefing for Claude Code: it sweeps the channels you follow, filters noise in two stages (title, then transcript), gives you a short briefing of the rest in your language in the chat, and adapts to your ratings — one video at a time.",
   "type": "module",
   "bin": {
     "yt-briefing": "dist/cli.js"

package/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "yt-briefing",
-  "version": "1.2.5",
-  "description": "The /yt briefing loop as a Claude Code pane: summary, one-key rating, research hand-off.",
+  "version": "1.3.0",
+  "description": "The /yt briefing loop in the chat: summary, rating dialog, research hand-off.",
   "author": {
     "name": "Michal Ryzio"
   },

package/plugin/hooks/engine.ts

@@ -1,4 +1,4 @@
-// How the pane runs the engine, relative to the project root (the session's working directory).
+// How the mod runs the engine, relative to the project root (the session's working directory).
 // This is the dev-clone form; `yt-briefing install-skill` / `init` rewrites it for the project it
 // installs into (the compiled dist/ under node_modules, with the runtime that ran the installer).
 export const engine = (name: string): string[] => ['bun', `src/${name}.ts`]

package/plugin/hooks/register.tsx

@@ -1,38 +1,23 @@
-import { atom, read, update } from 'claude-code'
 import type { EngineInterface, Register } from 'claude-code'
 
-import type { Pending, View } from '../types'
+import type { Pending } from '../types'
 import { engine } from './engine.ts'
 
-// The /yt rating loop as a pane. Every step runs the engine as a subprocess and reads its one-line
-// JSON; the model is not involved unless the person asks for research, which hands the video to
-// the session as a prompt. Ratings and comments are written by the engine, never by this module.
-// Where no pane can be seen (a Remote Control web client, a surface that places none) the same
-// loop runs in the engine's own question dialog instead.
-
-const PANE = 'yt-briefing'
-const view = atom({ plugin: 'yt-briefing', key: 'view' } as const, { phase: 'idle' } as View)
+// The /yt rating loop, in the chat on every surface. Each briefing is the model's whole answer to a
+// prompt this plugin submits, so the summary shows in full in the terminal, the desktop app and on
+// a phone; the rating dialog opens once that turn has ended. Ratings and comments are written by
+// the engine, never by this module.
 
 /** A sweep can expand channels, fetch transcripts and summarize; give it the engine's full budget. */
 const SWEEP_MS = 10 * 60_000
 /** A rating is a file write; a raw comment adds one `claude -p` call to distill it. */
 const RATE_MS = 3 * 60_000
 
-type Skip = { channel: string; title: string; reason: string }
 type SweepOut = {
   status: string
   summary?: string
   pending?: Pending
   lang?: string
-  error?: string
-  skipped?: number
-  skips?: Skip[]
-}
-
-const STATUS_TEXT: Record<string, string> = {
-  done: 'Sweep finished: nothing left to rate.',
-  rate_limited: 'YouTube is blocking this IP (429 / captcha). See README → Proxy.',
-  tooling_error: 'Transcript tooling failed (proxy, yt-dlp or network). See README → Proxy.',
 }
 
 /** The engine prints one JSON line on stdout; take the last non-empty line to be safe. */
@@ -41,64 +26,12 @@ function parseOut(stdout: string): SweepOut {
   return JSON.parse(line) as SweepOut
 }
 
-/** Replace the whole view; the one writer for a fresh phase. */
-async function setView($: EngineInterface, next: View) {
-  await update($, view, () => next)
-}
-
-function skipLine(out: SweepOut): string | undefined {
-  if (!out.skipped || !out.skips?.length) return undefined
-  const list = out.skips.map(s => `${s.channel} «${s.title}» — ${s.reason}`).join('; ')
-  return `Skipped ${out.skipped}: ${list}`
-}
-
 /** One engine sweep: the next ratable video, or the end state. Throws when the engine does not answer. */
-async function runSweep($: EngineInterface, reset: boolean): Promise<SweepOut> {
-  const run = await $.process.run([...engine('yt-sweep'), ...(reset ? ['--reset'] : [])], { timeoutMs: SWEEP_MS })
+async function runSweep($: EngineInterface): Promise<SweepOut> {
+  const run = await $.process.run(engine('yt-sweep'), { timeoutMs: SWEEP_MS })
   return parseOut(run.stdout)
 }
 
-/** What a sweep that found nothing to rate says, skips included. */
-function endText(out: SweepOut): string {
-  const text = out.status === 'done' ? STATUS_TEXT.done : out.error ?? STATUS_TEXT[out.status] ?? `Engine status: ${out.status}`
-  const skipped = skipLine(out)
-  return skipped ? `${text}\n${skipped}` : text
-}
-
-/** Advance to the next ratable video and put it (or the end state) in the view. */
-async function sweep($: EngineInterface, reset: boolean) {
-  await setView($, { phase: 'loading' })
-  let out: SweepOut
-  try {
-    out = await runSweep($, reset)
-  } catch (err) {
-    await setView($, { phase: 'error', message: `The engine did not answer: ${String(err)}` })
-    return
-  }
-  const skipped = skipLine(out)
-  if (out.status === 'rating_needed' && out.summary && out.pending) {
-    await setView($, { phase: 'rating', summary: out.summary, pending: out.pending, lang: out.lang, skipped })
-  } else if (out.status === 'done') {
-    await setView($, { phase: 'done', message: STATUS_TEXT.done, skipped })
-  } else {
-    await setView($, { phase: 'error', message: out.error ?? STATUS_TEXT[out.status] ?? `Engine status: ${out.status}`, skipped })
-  }
-}
-
-/** Record a rating (engine args), then move on; on failure stay on the video and say why. */
-async function rate($: EngineInterface, args: string[], thenNext = true): Promise<{ rule?: string } | null> {
-  const v = await read($, view)
-  if (v.phase !== 'rating' || v.busy) return null
-  await update($, view, cur => ({ ...cur, busy: args[0] === '--raw-comment' ? 'Turning the comment into a rule…' : 'Saving…' }))
-  const result = await record($, args)
-  if (!result) {
-    await update($, view, cur => ({ ...cur, busy: undefined }))
-    return null
-  }
-  if (thenNext) void sweep($, false)
-  return result
-}
-
 /** Run the rating engine; on failure toast why and resolve null. */
 async function record($: EngineInterface, args: string[]): Promise<{ rule?: string } | null> {
   const run = await $.process.run([...engine('yt-rating'), ...args], { timeoutMs: RATE_MS }).catch(err => ({
@@ -113,15 +46,6 @@ async function record($: EngineInterface, args: string[]): Promise<{ rule?: stri
   return result
 }
 
-/** Research: mark the video seen, close the pane and hand the video to the session. */
-async function research($: EngineInterface, question?: string) {
-  const v = await read($, view)
-  if (!v.pending || !v.summary) return
-  if (!(await rate($, ['--rating', '1'], false))) return
-  await $.ui.close({ id: PANE })
-  await handOff($, v.pending, v.summary, v.lang, question)
-}
-
 /** Give the video to the session as the person's prompt, to research it together. */
 async function handOff($: EngineInterface, p: Pending, summary: string, lang: string | undefined, question?: string) {
   const transcript = [...engine('yt-transcript'), p.videoId, '--lang', 'auto'].join(' ')
@@ -144,25 +68,8 @@ async function handOff($: EngineInterface, p: Pending, summary: string, lang: st
   }).catch(err => $.ui.toast(`Could not hand the video to the session: ${String(err)}`))
 }
 
-async function comment($: EngineInterface, raw: string) {
-  const text = raw.trim()
-  if (!text) return
-  if (text.toLowerCase() === 'stop') return void $.ui.close({ id: PANE })
-  if (text.startsWith('?')) return research($, text.slice(1).trim() || undefined)
-  const saved = await rate($, ['--raw-comment', text])
-  if (saved?.rule) $.ui.toast(`Rule saved: ${saved.rule}`)
-}
-
-/** True when the person typed /yt through Remote Control and nothing attached can show a pane. */
-async function paneUnseen($: EngineInterface, origin: { kind: string } | undefined): Promise<boolean> {
-  if (origin?.kind !== 'bridge') return false
-  return !(await $.session.surfaces()).some(s => s !== 'terminal')
-}
-
-// Where no pane can be seen the loop runs in the chat. A phone or web client draws a plugin's
-// command output and its dialogs, but not its log lines, and it folds text written mid-turn into a
-// one-line digest; only a turn's last message shows whole. So each briefing is the model's whole
-// answer to a prompt this plugin submits, and the rating dialog opens once that turn has ended.
+// A phone or web client draws a plugin's command output and its dialogs, but not its log lines,
+// and it folds text written mid-turn into a one-line digest; only a turn's last message shows whole.
 
 /** True while the chat loop runs; a prompt the person types ends it. */
 let looping = false
@@ -201,7 +108,7 @@ function submitLater($: EngineInterface, reset: boolean) {
 async function rateAfterTurn($: EngineInterface) {
   let out: SweepOut
   try {
-    out = await runSweep($, false)
+    out = await runSweep($)
   } catch {
     looping = false
     return
@@ -246,25 +153,17 @@ export const register: Register = on => {
   on('session.start', async ($, e, next) => {
     await $.command.register({
       name: 'yt',
-      description: 'Rate the next videos from the YouTube channels you follow, in a pane',
+      description: 'Rate the next videos from the YouTube channels you follow',
     })
 
     return next(e)
   })
 
-  on('command.run', { command: 'yt' }, async ($, e) => {
-    if (!(await paneUnseen($, e.origin))) {
-      const opened = await $.ui.open({ id: PANE, title: 'yt-briefing', focus: true, closeOnEscape: true })
-      if (opened.isPlaced) {
-        void sweep($, true)
-        return { text: 'Briefing opened in a pane.' }
-      }
-      await $.ui.close({ id: PANE })
-    }
+  on('command.run', { command: 'yt' }, async ($) => {
     looping = true
     submitLater($, true)
 
-    return { text: 'No pane here: the briefing runs in the chat.' }
+    return { text: 'The briefing runs in the chat.' }
   })
 
   // The engine shows a plugin none of its own prompts, so a prompt seen here is the person's.
@@ -284,57 +183,4 @@ export const register: Register = on => {
     if (looping && !e.agentId && e.reason === 'answer') void rateAfterTurn($)
     return next(e)
   })
-
-  on('ui.render', { component: 'Pane', requestId: PANE }, async ($, e) => {
-    const { Box, Text, Button, Markdown } = $.ui.resolve(e)
-    const v = await read($, view)
-    const close = () => void $.ui.close({ id: PANE })
-
-    if (v.phase === 'idle' || v.phase === 'loading') {
-      return <Text dimColor>Finding the next video…</Text>
-    }
-
-    if (v.phase !== 'rating') {
-      return (
-        <Box flexDirection="column">
-          {v.skipped && <Text dimColor>{v.skipped}</Text>}
-          <Text>{v.message ?? ''}</Text>
-          <Button key="close" label="Close" hotkey="c" role="dismiss" onPress={close} />
-        </Box>
-      )
-    }
-
-    const idle = !v.busy
-    const buttons = (
-      <Box flexDirection="row" gap={2}>
-        <Button key="ok" label="OK" hotkey="1" variant="primary" dimColor={!idle}
-          onPress={() => void rate($, ['--rating', '1'])} />
-        <Button key="weak" label="Weak" hotkey="2" dimColor={!idle}
-          onPress={() => void rate($, ['--rating', '0'])} />
-        <Button key="research" label="Research" hotkey="3" dimColor={!idle}
-          onPress={() => void research($)} />
-        <Button key="stop" label="Stop" hotkey="4" role="dismiss" onPress={close} />
-      </Box>
-    )
-
-    let field = null
-    if (e.surface !== 'mobile') {
-      const { Input } = $.ui.resolve(e)
-      field = (
-        <Input key="comment" label="Comment" submitLabel="save rule"
-          placeholder="becomes a rule for this channel · ?question = research · stop"
-          onSubmit={value => void comment($, value)} />
-      )
-    }
-
-    return (
-      <Box flexDirection="column" gap={1}>
-        {v.skipped && <Text dimColor>{v.skipped}</Text>}
-        <Markdown text={(v.summary ?? '').slice(0, 10000)} />
-        {v.busy ? <Text dimColor>{v.busy}</Text> : buttons}
-        {field}
-        <Text dimColor>OK = neutral · Weak = teach the filter to skip titles like this · Research = dig into it with Claude</Text>
-      </Box>
-    )
-  })
 }

package/plugin/types/index.d.ts

@@ -5,24 +5,3 @@ export type Pending = {
   type: string
   publishedAt: string
 }
-
-export type View = {
-  phase: 'idle' | 'loading' | 'rating' | 'done' | 'error'
-  /** Briefing markdown of the video awaiting a rating. */
-  summary?: string
-  pending?: Pending
-  /** Language the engine writes in; the research hand-off asks for it too. */
-  lang?: string
-  /** Videos the filters dropped on the way to this one, one line. */
-  skipped?: string
-  /** Error or status text for the done/error phases. */
-  message?: string
-  /** Set while a rating or comment is being written; presses are ignored meanwhile. */
-  busy?: string
-}
-
-declare module 'claude-code' {
-  interface PluginState {
-    'yt-briefing': { view: View }
-  }
-}