youtube-transcripts

youtube-transcripts is a skill for Claude Code, Codex from haljishi/vidwords-mcp. It costs 111 tokens per session (1,373 once invoked), scanned A, original, MIT.

A tool for reading and searching YouTube transcripts through VidWords. A transcript is the written text of what is spoken in a video, with timestamps that show when each part occurs.

In plain words
What is it for?
Use it to find what a video says about a topic, retrieve subtitles as TXT, SRT, or VTT, or provide quotes with clickable timestamps.
Why use it?
It lets an agent answer questions about a video without watching it and link answers to the relevant moments.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit Use it to find what a video says about a topic, retrieve subtitles as TXT, SRT, or VTT, or provide quotes with clickable timestamps.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/haljishi/vidwords-mcp/youtube-transcripts
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add haljishi/vidwords-mcp --skill youtube-transcripts
Clone the repo
git clone --depth 1 https://github.com/haljishi/vidwords-mcp

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for youtube-transcripts

README.md
[![agentmods](https://agentmods.dev/badge/skills/haljishi/vidwords-mcp/youtube-transcripts/github.svg)](https://agentmods.dev/skills/haljishi/vidwords-mcp/youtube-transcripts)
Your own site
<a href="https://agentmods.dev/skills/haljishi/vidwords-mcp/youtube-transcripts"><img src="https://agentmods.dev/badge/skills/haljishi/vidwords-mcp/youtube-transcripts/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for youtube-transcripts

Your own site · 80×15
<a href="https://agentmods.dev/skills/haljishi/vidwords-mcp/youtube-transcripts"><img src="https://agentmods.dev/badge/skills/haljishi/vidwords-mcp/youtube-transcripts.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 111 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,373 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00111 $0.01373
Opus 5 $0.00056 $0.00687
Sonnet 5 $0.00022 $0.00275
Haiku 4.5 $0.00011 $0.00137

Measured 8d ago against content hash e3e04a5a31e4, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

youtube-transcripts scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/youtube-transcripts/SKILL.md · 128 lines

How it starts

The opening of the file, as written. The whole thing — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Reading YouTube videos

A model cannot watch a video, but it can read a transcript carrying timestamps it is able to cite. So the useful unit here is not "a transcript" — it is a grounded answer with a link the user can click and check.

Prefer returning a quote plus its deep link over a wall of text.

Endpoint

Hosted MCP server, no integration code:

POST https://vidwords.com/mcp
Authorization: Basic <api-token>

Basic, not Bearer, and the token is not a base64 pair — send it verbatim. Get one from https://vidwords.com/profile. Free tier included; the account's email must be verified or every call returns 403.

There is a plain REST equivalent at POST https://vidwords.com/api/transcripts if MCP is not available to you — see https://vidwords.com/api-docs.

Choosing a tool

search_transcript is the default. Reach for it whenever the question is "what does this video say about X".

{ "video": "dQw4w9WgXcQ", "query": "pricing model" }

It returns the matching moments, each with timestamp, startSeconds and a youtube.com/watch?v=…&t=…s url. Cite those urls.

get_transcript returns the whole text and costs the same. A two-hour interview is ~20,000 words of which perhaps 300 answer the question, and the other 19,700 compete for your attention and degrade the answer. Use it only when the full text is genuinely the deliverable: an export, a diff, a corpus.

{ "videos": ["dQw4w9WgXcQ", "9bZkp7q19f0"], "lang": "en" }

Ask for a span, not a whole video

Both tools take optional from/to timecodes. Accepted forms: seconds (615), m:ss (10:20), h:mm:ss (1:02:13) — the same formats the tools print back, so a timestamp from one answer can be pasted into the next question.

{ "video": "dQw4w9WgXcQ", "query": "revenue", "from": "10:20", "to": "11:00" }

When the user names a time span, pass it rather than fetching everything and filtering yourself. It is the same credit either way, but it keeps your context clear and the reply focused.

Read the full file on GitHub · 128 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 128 lines · 0 tokens per session scan A e3e04a5a31e4

Subscribe to this mod's changes

youtube-transcripts is a skill published in the GitHub repository haljishi/vidwords-mcp (0 stars, last pushed 17d ago), licensed MIT. It adds 111 tokens to every session and 1,373 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.