CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

watch

Watch a video (URL or local path). Downloads with yt-dlp, extracts auto-scaled frames with ffmpeg, pulls the transcript from captions (or local WhisperX / cloud Whisper fallback), and hands the result to the agent so it can answer questions about what's in the video. With a Gemini API key, Google's agentic video model watches the full video instead.

Quality

84%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

88%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

An exceptionally actionable, well-sequenced skill body with explicit validation and error-recovery loops throughout the setup and watch flows. Its main costs are length and inlining: version- and backend-specific detail that would fit better in one-level-deep reference files sits in the main body.

Suggestions

Move the WhisperX configuration defaults, chunking budgets, and model/timeout tuning into a references/ file (e.g. TRANSCRIPTION.md), keeping a one-line pointer in SKILL.md.

Trim the field-by-field enumeration of `setup.py --json` output to the fields the agent must act on (engine, can_proceed, first_run), leaving the rest to the script's own output.

Relocate version-sensitive notes (WhisperX 3.8.6 language mislabeling, gemini-3.7-flash default) to a reference file so the main body does not age.

DimensionReasoningScore

Conciseness

The body is dense, imperative, and almost entirely skill-specific operational detail Claude cannot infer (exit codes, env-var precedence, caps, backend behavior), with no padding explaining known concepts. Minor instances could be trimmed — e.g., the Windows/PowerShell interpreter paragraph, the full field-by-field enumeration of `setup.py --json` output, and version-specific notes like 'WhisperX 3.8.6 mislabels its JSON language' — placing it at the 'efficient with minor over-explanation' anchor rather than the every-token-earns-its-place anchor at 5.

4 / 5

Actionability

Commands are copy-paste ready ('python3 "${SKILL_DIR}/scripts/setup.py" --json', the quoted watch.py invocation, rerun flows for each exit code), and the option table gives concrete defaults (`--resolution W` default 512, `--fps` at most 2, SS/MM:SS formats). This matches the 'fully executable; copy-paste ready; specific examples cover the common cases' anchor, above the minor-gaps anchor at 4.

5 / 5

Workflow Clarity

The sequence is explicit — setup preflight → engine question → key handling (exit 3 → paste or open editor → rerun → exit 0) → detail/backend questions → watch → read frames and answer — with explicit validation checkpoints and feedback loops at every fragile step (exit codes, 'No silent fallback', 403 → upgrade yt-dlp → retry once, failed-install never auto-switches backends). This matches the 'clear sequence with explicit validation steps; feedback loops for error recovery' anchor.

5 / 5

Progressive Disclosure

Sections are well organized and every referenced bundled script exists in scripts/ (watch.py, setup.py, download.py, frames.py, transcribe.py, whisper.py, local_whisperx.py, gemini.py, config.py, runtime.py — all verified present), with references one level deep. However, substantial specialist detail (WhisperX configuration, cloud chunking budgets, the security/runtime-access section) is inlined in SKILL.md rather than split into reference files, so it sits at the 'good structure; minor organization gaps' anchor rather than the 'content appropriately split' anchor at 5.

4 / 5

Total

18

/

20

Passed

Description

75%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A highly specific, third-person description that names concrete tools and the full pipeline, with an unusual degree of operational detail. Its main weakness is the absence of an explicit 'Use when...' trigger clause, which caps completeness, and a few missing natural synonyms like 'YouTube' or 'transcribe'.

Suggestions

Add an explicit trigger clause, e.g. 'Use when the user shares a video URL or file and asks what's in it, wants a summary, or asks about something shown at a timestamp.'

Include common natural synonyms users say, such as 'YouTube link', 'video transcript', or 'analyze this video'.

Trim internal engine detail (e.g., the WhisperX/cloud fallback ordering) to leave room for the trigger clause without lengthening the description.

DimensionReasoningScore

Specificity

The description lists multiple concrete, tool-named actions — 'Downloads with yt-dlp, extracts auto-scaled frames with ffmpeg, pulls the transcript from captions (or local WhisperX / cloud Whisper fallback), and hands the result to the agent' — with comprehensive coverage of the pipeline plus the Gemini alternative. This matches the 'multiple specific concrete actions; comprehensive coverage' anchor, not the score-4 anchor which expects minor coverage gaps.

5 / 5

Completeness

The 'what' is clear and detailed, but there is no 'Use when...' clause or equivalent explicit trigger guidance — 'so it can answer questions about what's in the video' only weakly implies when to invoke the skill. Per the rubric guideline, a missing explicit trigger clause caps completeness at 3; it is above 2 because the 'what' is fully stated.

3 / 5

Trigger Term Quality

'Watch a video (URL or local path)', 'transcript', 'captions', and "answer questions about what's in the video" are natural phrases a user would say, giving good keyword coverage. A few common variations are missing (e.g., 'YouTube link', 'analyze video', 'transcribe'), so it falls between the good (4) and comprehensive (5) anchors, closer to 4.

4 / 5

Distinctiveness Conflict Risk

The video-watching niche with named tools (yt-dlp, ffmpeg, WhisperX, Gemini video model) is clearly distinct with minimal overlap risk against generic media or file skills, matching the 'clear niche with distinct triggers' anchor rather than the minor-overlap anchor at 4.

5 / 5

Total

17

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

Total

15

/

16

Passed

Repository

bradautomates/claude-video

Path

skills/watch/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.