Claude Watch

What it does

When your agent uses it

• Tasks that involve Study guides and flashcards
• Tasks that involve Video production
• Tasks that involve Note-taking

Example prompts

• “/claude-watch”

Requirements

• Python 3
• A Bash shell
• Pre-approved tools (allowed-tools): Bash, Read, Write, AskUserQuestion

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

1. 1Title and slug
2. 2Number of sections + key concepts
3. 3Path to the notes file

What it can do on your machine

Read from SKILL.md and the folder at commit ade7d61. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Bash
  • Read
  • Write
  • AskUserQuestion

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Ships 1 file in scripts/ (Python and Shell, from the files we listed), which the agent can run.

  Shell commands in SKILL.md call:

  • python3
  • brew

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Claude Watch loads about 1.6k tokens when it runs. Until then it costs about 85 tokens; SKILL.md has 678 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~85

When it runs · the whole SKILL.md, loaded when a task matches

~1.6k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

• NoteMentions a .env fileSKILL.md:30

  ds. It scaffolds `~/.config/claude-watch/.env` (mode 0600) with commented placeholders.

• NoteMentions a .env fileSKILL.md:32

  and write it to `~/.config/claude-watch/.env`. If they don't want to, run with `--no-whisper`; videos without native ca

• NoteMentions a .env fileSKILL.md:146

  - Reads/writes `~/.config/claude-watch/.env` (mode 0600) for keys

• NotePre-approves every shell command (allowed-tools: Bash)SKILL.md

  allowed-tools: Bash, Read, Write, AskUserQuestion

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from devinilabs/claude-watch at commit ade7d61, republished under its MIT licence (© devinilabs). 678 words, ~1,639 tokens.

name

claude-watch

description

Watch a tutorial or lecture video (URL or local path) and produce structured study notes. Downloads with yt-dlp, detects scene changes with ffmpeg, pulls a timestamped transcript (captions or Whisper API fallback), and writes a section-by-section markdown notes file with embedded screenshots to ~/claude-watch/library/<slug>/.

allowed-tools

Bash, Read, Write, AskUserQuestion

argument-hint

<video-url-or-path> [topic-or-question]

homepage

https://github.com/devinilabs/claude-watch

repository

https://github.com/devinilabs/claude-watch

license

MIT

user-invocable

true

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

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

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

# <Video Title>

**Source:** <URL or path>  ·  **Duration:** MM:SS  ·  **Watched:** YYYY-MM-DD

## TLDR
<3-4 sentences: what the video is about and the single most important takeaway.>

## Key Concepts
- **<concept>** — <one-line definition> · `[t=MM:SS]`
- ...

## Notes

### [t=00:04] <Section title you derive from on-screen + spoken content>

![](frames/0001_t00-04.jpg)

**On screen:** <Transcribe / describe the slide, code, diagram. If code, transcribe verbatim.>

**Said:** <Relevant transcript excerpt for this scene, lightly cleaned.>

**Synthesis:** <Your connection — what this section is teaching, how it links to prior section.>

### [t=00:31] <next section>
...

## Code & Commands
<every code-on-screen frame's content as a runnable fenced block, language-tagged, with [t=MM:SS] back-link>

```python
# [t=03:45]
def forward(x):
    return x @ W + b
```

## Diagrams Referenced
- `[t=02:10]` — <one-line description of the diagram in frame 0008>
- ...

## Open Questions
- <things mentioned but not fully covered, or follow-ups to explore>

© devinilabs, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 40 other files (scripts) in the repository root of devinilabs/claude-watch.

• SKILL.md
• .claude-plugin/marketplace.json
• .claude-plugin/plugin.json
• .codex-plugin/skill.json
• .gitattributes
• .github/workflows/release.yml
• .gitignore
• CHANGELOG.md
• LICENSE
• README.md
• commands/claude-watch.md
• conftest.py
• hooks/hooks.json
• hooks/scripts/session_start.sh
• … and 27 more

Open the folder on GitHubat commit ade7d61