Watch

/watch

You don't have a video input; this skill gives you one. A Python script gets captions first, optionally downloads the video, extracts frames as JPEGs (scene-aware, or fast keyframes at efficient detail), gets a timestamped transcript (native captions first, then Deepgram as fallback), and prints frame paths. You then Read each frame path to see the images and combine them with the transcript to answer the user.

Resolve SKILL_DIR (do this before any command)

Every python3 ... command below runs a bundled script under SKILL_DIR/scripts/. Set SKILL_DIR to the absolute path of the directory containing THIS SKILL.md you just Read — your harness told you that path in the Read result. The scripts are always a direct sibling of this file (SKILL_DIR/scripts/watch.py), in every install layout:

Read ~/.claude/plugins/cache/epictaste/watch/<ver>/skills/watch/SKILL.md → SKILL_DIR=…/skills/watch
Read ~/.codex/skills/watch/SKILL.md                                          → SKILL_DIR=~/.codex/skills/watch
Read ~/.agents/skills/watch/SKILL.md                                         → SKILL_DIR=~/.agents/skills/watch

Substitute that literal path for ${SKILL_DIR} in every command. This works on every harness (Claude Code, Codex, Cursor, Gemini CLI, …) without relying on any harness-specific environment variable. Guard once at the start of a run:

SKILL_DIR="<absolute path of the directory containing the SKILL.md you Read>"
if [ ! -f "$SKILL_DIR/scripts/watch.py" ]; then
  echo "ERROR: scripts/watch.py not found under SKILL_DIR=$SKILL_DIR" >&2
  echo "Re-check the directory of the SKILL.md you Read and substitute it as SKILL_DIR." >&2
  exit 1
fi

Step 0 — Setup preflight (runs every /watch invocation, silent on success)

Python interpreter: every python3 ... command in this skill is for macOS/Linux. On Windows, substitute python — the python3 command on Windows is the Microsoft Store stub and will not run the script.

On the first /watch invocation in a session, use structured preflight so you can detect first-run setup:

python3 "${SKILL_DIR}/scripts/setup.py" --json

Branch on two fields:

• can_proceed: true and first_run: false → setup is already done (the user may have deliberately skipped a Deepgram key — that's allowed). Proceed to Step 1 without comment.
• first_run: true → genuine first-time setup. Do these in order:
  1. If missing_binaries is non-empty, run the installer first (it auto-installs on macOS / prints commands elsewhere — see below) and confirm the binaries land. Do not skip this and jump to preferences.
  2. Run the installer once more if needed so it scaffolds ~/.config/watch/.env (it only writes the template when the file is absent, so let it create the file before you write any values into it).
  3. Encourage a Deepgram API key and ask the watch-preference questions below, then write the selected values into ~/.config/watch/.env and set SETUP_COMPLETE=true.
• can_proceed: false and first_run: false → setup was finished before but the environment regressed (e.g. missing_binaries after an OS change). Run the installer to remediate, then proceed. Don't re-ask preferences.

A missing Deepgram key is encouraged to fix, not required: on a genuine first run status will read needs_key even when binaries are present — that's your cue to encourage a key, not a blocker.

On follow-up /watch calls in the same session, use the silent check:

python3 "${SKILL_DIR}/scripts/setup.py" --check

This is a <100ms lookup. Exit 0 means /watch can run — this includes a user who finished setup without a Deepgram key (keyless is allowed). On exit 0 the script emits nothing — proceed to Step 1 without comment. Do NOT announce "setup is complete" to the user — they don't need a status message on every turn. The only acceptable user-visible output from Step 0 is when remediation is required.

On non-zero exit, follow the table:

Exit	Meaning	Action
2	Missing binaries (ffmpeg / ffprobe / yt-dlp)	Run installer
3	Genuine first run with no Deepgram API key	Run installer to scaffold .env, then encourage a key (the user may decline — proceed with --no-transcribe)
4	Both missing	Run installer, then encourage a key

Exit 3 only fires before the user has completed setup. Once SETUP_COMPLETE=true is written, a keyless install returns exit 0 and is never nagged again.

The installer is idempotent — safe to re-run:

python3 "${SKILL_DIR}/scripts/setup.py"

On macOS with Homebrew, it auto-installs ffmpeg and yt-dlp. On Linux/Windows, it prints the exact install commands for the user to run. It scaffolds ~/.config/watch/.env with commented placeholders and default watch settings at 0600 perms.

If an API key is still missing after install: encourage a Deepgram API key (nova-3 handles German + English with precise timestamps; new accounts get $200 free credit — https://console.deepgram.com). Write it into ~/.config/watch/.env — set the DEEPGRAM_API_KEY=... line. If they don't want to set up transcription, proceed with --no-transcribe and tell them videos without native captions will come back frames-only.

First-run watch preference: after the installer has scaffolded ~/.config/watch/.env, use AskUserQuestion to ask one question:

• Default detail (one dial). Present these as AskUserQuestion options in this exact order — lightest to heaviest — and keep (recommended) on balanced even though it is not first (do not reorder to put the recommended option first):
  • transcript — no frames at all, transcript only (skips video download when captions exist).
  • efficient — fast keyframe pass (cap 50).
  • balanced (recommended) — scene-aware frames (cap 100, default).
  • token-burner — scene-aware, uncapped (maximum fidelity; high token cost).

Write the answer directly into ~/.config/watch/.env by setting the bare key on its own line — no trailing inline comment (a # note after the value can break parsing):

WATCH_DETAIL=balanced

Use the user's selected value. If they skip the question, keep the recommended default. Once dependencies, the API-key choice, and this preference are handled, write or update SETUP_COMPLETE=true in the same file. Do not ask this preference question again when SETUP_COMPLETE=true.

Structured mode (optional): python3 "${SKILL_DIR}/scripts/setup.py" --json emits {status, can_proceed, first_run, setup_complete, missing_binaries, transcriber_backend, has_api_key, config_file, watch_detail, platform} where status is one of ready | needs_install | needs_key | needs_install_and_key. status describes the ideal state (a key is encouraged, so a keyless first run reads needs_key); can_proceed is the operational gate (binaries present AND a key is set OR setup was already completed). Branch on can_proceed/first_run to decide whether to run; use status to decide what to encourage.

Within a single session, you can skip Step 0 on follow-up /watch calls — once --check returned 0, nothing about the environment changes between turns.

When to use

• User pastes a video URL (YouTube, Vimeo, X, TikTok, Twitch clip, most yt-dlp-supported sites) and asks about it.
• User points at a local video file (.mp4, .mov, .mkv, .webm, etc.) and asks about it.
• User types /watch <url-or-path> [question].

Recommended limits

• Best accuracy: videos under 10 minutes. Frame coverage scales inversely with duration.
• Universal rate cap: 2 fps. The script never samples faster than 2 fps, even when a budget or --fps would imply more.
• The frame ceiling is set by the detail mode (WATCH_DETAIL in ~/.config/watch/.env, or --detail), not a single global cap:
  • transcript → no frames
  • efficient → up to 50 (keyframes)
  • balanced (default) → up to 100 (scene-aware)
  • token-burner → uncapped (scene-aware; a soft warning prints past 250 frames)
  • --max-frames N overrides whichever cap the mode would otherwise use.
• Full-video frame budget by duration. Token cost grows with frame count, so the script targets a budget by duration. This budget sets the fps and the uniform-sampling fallback; scene-aware selection can fill up to the detail cap above, whichever is lower:
  • ≤30s → ~12-30 frames
  • 30s-1min → ~40 frames
  • 1-3min → ~60 frames
  • 3-10min → ~80 frames
  • >10min → up to the detail cap, sparsely spaced (warning printed)
• If the user hands you a long video, consider asking whether they want a specific section before burning tokens on a sparse scan.

How to invoke

Step 1 — parse the user input. Separate the video source (URL or path) from any question the user asked. Example: /watch https://youtu.be/abc what language is this in? → source = https://youtu.be/abc, question = what language is this in?.

Step 2 — run the watch script. Pass the source verbatim. Do not shell-escape it yourself beyond normal quoting:

python3 "${SKILL_DIR}/scripts/watch.py" "<source>"

Optional flags:

• --detail transcript|efficient|balanced|token-burner — fidelity/speed dial. transcript = no frames (transcript only, skips video download when captions exist); efficient = fast keyframes (cap 50); balanced = scene-aware frames (cap 100); token-burner = scene-aware, uncapped.
• --start T / --end T — focus on a section. Accepts SS, MM:SS, or HH:MM:SS. When either is set, fps auto-scales denser (see "Focusing on a section" below).
• --timestamps T1,T2,… — grab a frame at each of these absolute timestamps (SS, MM:SS, or HH:MM:SS). Use this after reading the transcript to capture deictic moments the presenter flags ("look here", "as you can see", "notice this") that visual selection alone may miss. See "Transcript-cue frames" below.
• --max-frames N — override the preset cap for tighter token budget (e.g. --max-frames 40)
• --resolution W — change frame width in px (default 512; bump to 1024 only if the user needs to read on-screen text)
• --fps F — override auto-fps (clamped to 2 fps max)
• --out-dir DIR — keep working files somewhere specific (default: an auto-generated tmp dir)
• --no-transcribe — disable Deepgram transcription entirely (frames-only if no captions)
• --no-dedup — keep near-duplicate frames. By default a frame-delta pass drops frames that are visually near-identical to the previous kept one (held slides, static screen recordings, paused video) so the frame budget goes to distinct content; the report's Frames line notes how many were dropped. Pass this only if the user needs every sampled frame (e.g. judging subtle frame-to-frame motion).

Focusing on a section (higher frame rate)

When the user asks about a specific moment — "what happens at the 2 minute mark?", "zoom into 0:45 to 1:00", "the first 10 seconds" — pass --start and/or --end. The script switches to focused-mode budgets, which are denser than full-video budgets (still capped at 2 fps, and still bounded by the detail-mode cap — the counts below assume the default balanced cap of 100; efficient tops out at 50):

• ≤5s → 2 fps (up to 10 frames)
• 5-15s → 2 fps (up to 30 frames)
• 15-30s → ~2 fps (up to 60 frames)
• 30-60s → ~1.3 fps (up to 80 frames)
• 60-180s → ~0.6 fps (100 frames, capped)

Focused mode is the right call for:

• Any moment/range the user names explicitly ("around 2:30", "the intro", "the last 30 seconds").
• Any video longer than ~10 minutes where the user's question is about a specific part — running focused on the relevant section is far more useful than a sparse scan of the whole thing.
• Re-runs after a full scan didn't have enough detail in some region.

Transcript is auto-filtered to the same range. Frame timestamps are absolute (real video timeline, not offset-from-start).

Examples:

# Last 10 seconds of a 1 minute video
python3 "${SKILL_DIR}/scripts/watch.py" video.mp4 --start 50 --end 60

# Zoom into 2:15 → 2:45 at 2 fps (60 frames)
python3 "${SKILL_DIR}/scripts/watch.py" "$URL" --start 2:15 --end 2:45 --fps 2

# From 1h12m to the end of the video
python3 "${SKILL_DIR}/scripts/watch.py" "$URL" --start 1:12:00

Step 3 — Read every frame path the script lists. The Read tool renders JPEGs directly as images for you. Read all frames in a single message (parallel tool calls) so you see them together. The frames are in chronological order with a t=MM:SS timestamp so you can align them to the transcript.

Step 4 — answer the user. You now have two streams of evidence:

• Frames — what's on screen at each timestamp
• Transcript — what's said at each timestamp. The report's header shows the source (captions = yt-dlp pulled native subs; deepgram = transcribed by the Deepgram API).

If the user asked a specific question, answer it directly citing timestamps. If they didn't ask anything, summarize what happens in the video — structure, key moments, notable visuals, spoken content.

This holds for transcript detail too: even with no frames, produce a summary like the other modes — do not paste the full transcript into chat. Synthesize structure, key moments, and spoken content with timestamps; quote only the lines that matter. Offer the raw transcript only if the user explicitly asks for it.

Step 5 — clean up. The script prints a working directory at the end. If the user isn't going to ask follow-ups about this video, delete it with rm -rf <dir>. If they might, leave it in place.

Detail and frames

Default behavior comes from ~/.config/watch/.env:

• WATCH_DETAIL=transcript|efficient|balanced|token-burner (default: balanced)

At transcript detail, captions are enough to return a report without downloading video. If captions are missing, the script downloads audio only and tries Deepgram. If no transcript can be produced, it reports the limitation clearly; re-run with --detail balanced for frames.

At efficient detail, the script downloads the video and extracts keyframes only (ffmpeg -skip_frame nokey) — a near-instant pass that lands frames on scene cuts. If a clip has fewer than 4 keyframes it falls back to uniform sampling.

At balanced / token-burner detail, the script extracts scene-aware frames: ffmpeg scene-change selection first, falling back to uniform sampling only when the video is effectively static. balanced caps at 100 frames; token-burner is uncapped. Frame report lines include both timestamp and selection reason. Extracted images are clamped to a maximum 1998px height for Claude Read compatibility.

Transcript-cue frames

Visual frame selection (scene/keyframe) can miss the moments a presenter explicitly flags — "look here", "as you can see", "notice this", "watch what happens" — because pointing at a slide is often a low visual change. --timestamps lets you force a frame at those exact moments. You decide which moments matter, by reading the transcript:

1. Run once at --detail transcript (or any detail) to get the timestamped transcript.
2. Scan it for deictic cues — phrases where the speaker directs attention to something on screen. This is a judgment call (ignore rhetorical "look, the point is…"); that's why it's done by you, not a regex.
3. Re-run with --timestamps 4:32,7:10,9:55 (absolute source times). For a URL, point the second run at the downloaded local file in the work dir so it doesn't re-download.

Behavior:

• Additive by default. Cue frames (reason=transcript-cue) are merged into whatever --detail already selected, in chronological order.
• Pinned and counted first. Cue frames are reserved against the frame cap before the detail engine runs, so they're never evicted by even-sampling.
• Honors focus mode. With --start/--end, any cue timestamp outside the window is dropped (reported in the summary). Coordinates are always absolute source time.
• Cue-only frames. --detail transcript --timestamps … skips scene/keyframe sampling and returns only the cue frames (it will download the video to do so, since frames need pixels).

Transcription

The script gets a timestamped transcript in one of two ways:

1. Native captions (free, preferred). yt-dlp pulls manual or auto-generated subtitles from the source platform if available.
2. Deepgram fallback. If no captions came back (or the source is a local file), the script extracts audio (ffmpeg -vn -ac 1 -ar 16000 -b:a 64k, ~0.5 MB/min) and POSTs the raw bytes to Deepgram's /v1/listen endpoint. It uses model=nova-3 with language=multi (handles German, English, and mixed clips) and utterances=true for sentence-level timestamps. Deepgram's 2 GB upload cap means our tiny audio never needs chunking.

The DEEPGRAM_API_KEY lives in ~/.config/watch/.env (get one at https://console.deepgram.com — new accounts get $200 free credit). Optional overrides in the same file: WATCH_DEEPGRAM_MODEL (default nova-3) and WATCH_DEEPGRAM_LANGUAGE (default multi; set de or en for slightly better accuracy and lower cost when you know the language). Use --no-transcribe to skip transcription entirely.

Cookies (login-gated sites: Instagram, TikTok, X)

Public videos (YouTube, Vimeo, most sites) need nothing. But Instagram, TikTok, and X reject anonymous downloads — yt-dlp returns an "empty media response" or a login error. To watch those, set one cookie source in ~/.config/watch/.env (default is none):

• WATCH_COOKIES_FILE=~/.config/watch/cookies.txt — robust and platform-stable. Export once with a browser extension like "Get cookies.txt LOCALLY". On modern macOS Chrome this is the reliable path (Chrome's App-Bound Encryption often breaks the browser method below).
• WATCH_COOKIES_FROM_BROWSER=chrome:Profile 3 — convenient, reads live from the named browser profile. Format is browser[:profile]. Handy but can fail on recent macOS Chrome ("could not decrypt cookie"); Firefox is a reliable alternative there.

Cookies are a full login session — never commit them (.env and cookies.txt are gitignored). If a download fails with an auth/empty-response error on a gated site, tell the user this is why and point them at these settings.

Cookies are never sent to YouTube, even when configured. Public YouTube needs none, and passing them forces yt-dlp onto its tv/web client, which requires solving the JS "n-challenge" (a deno/node runtime) — without one, the download dies with "No video formats found". Skipping cookies keeps YouTube on the cookie-free android_vr client, which works with zero extra dependencies.

Failure modes and handling

• Setup preflight failed → run python3 "${SKILL_DIR}/scripts/setup.py" (auto-installs ffmpeg/yt-dlp via brew on macOS, scaffolds the .env). For API key, ask the user via AskUserQuestion and write it to ~/.config/watch/.env.
• No transcript available → captions missing AND (no Deepgram key OR Deepgram API failed). Script prints a hint pointing to setup. Proceed frames-only and tell the user.
• Long video warning printed → acknowledge it in your answer. Offer to re-run focused on a specific section via --start/--end rather than a sparse full-video scan.
• Download fails on Instagram/TikTok/X → yt-dlp's "empty media response" or login error means the site needs cookies. Point the user at the Cookies section (WATCH_COOKIES_FILE / WATCH_COOKIES_FROM_BROWSER). For other login-required or region-locked videos, tell the user plainly; do not keep retrying.
• Deepgram request fails → the error is printed to stderr. A 401 means an invalid key (or the wrong auth scheme) — check DEEPGRAM_API_KEY. A 429 is a rate limit (the script retries with backoff). A 504 means the file was too long for Deepgram's ~10 min processing window — focus on a section with --start/--end. The report says "none available" if transcription fails entirely.

Token efficiency

This skill burns tokens primarily on frames. Order of magnitude:

• 80 frames at 512px wide is roughly 50-80k image tokens depending on aspect ratio.
• The transcript is cheap (a few thousand tokens at most for a 10-minute video).
• Bumping --resolution to 1024 roughly quadruples the image tokens per frame. Only do it when necessary.

If you already watched a video this session and the user asks a follow-up, do not re-run the script — you already have the frames and transcript in context. Just answer from what you have.

Security & Permissions

What this skill does:

• Runs yt-dlp locally to download the video and pull native captions when the source supports them (public data; the request goes directly to whatever host the URL points at)
• Runs ffmpeg / ffprobe locally to extract frames as JPEGs and, when transcription is needed, a mono 16 kHz audio clip
• Sends the extracted audio clip to Deepgram's API (api.deepgram.com/v1/listen) when DEEPGRAM_API_KEY is set and native captions are missing
• When a cookie source is configured (WATCH_COOKIES_FILE / WATCH_COOKIES_FROM_BROWSER), passes those login cookies to yt-dlp so it can fetch login-gated videos (Instagram, TikTok, X). Cookies never leave the local yt-dlp process.
• Writes the downloaded video, frames, audio, and an intermediate transcript to a working directory under the system temp dir (or --out-dir if specified) so Claude can Read them
• Reads / creates ~/.config/watch/.env (mode 0600) to store the Deepgram API key and a SETUP_COMPLETE marker. As a fallback, also reads .env in the current working directory

What this skill does NOT do:

• Does not upload the video itself to any API — only the extracted audio goes out, and only when native captions are missing AND transcription is not disabled with --no-transcribe
• Does not access any platform account unless you explicitly configure a cookie source — with no cookies set, yt-dlp only ever requests public data
• Does not log, cache, or write the API key or cookies to stdout, stderr, or output files
• Does not persist anything outside the working directory and ~/.config/watch/.env — clean up the working directory when you're done (Step 5)

Bundled scripts: scripts/watch.py (entry point), scripts/download.py (yt-dlp wrapper + cookies), scripts/frames.py (ffmpeg frame extraction), scripts/transcribe.py (caption parsing), scripts/deepgram.py (Deepgram client), scripts/setup.py (preflight + installer)

Review scripts before first use to verify behavior.