transcript-extraction

A research skill for extracting and cleaning transcripts from video sources, mainly YouTube. It turns captions or speech into readable text for later searching and analysis.

In plain words
What is it for?
Getting captions or auto-generated transcripts, cleaning them, and saving readable transcript files for research.
Why use it?
It removes timestamps, speaker markers, and repeated filler so video-based research is easier to read and reference.

Skill for Claude CodeCodex

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.

agentmods
npx agentmods add skills/endogenai/dogma/transcript-extraction
Any agent
npx skills add EndogenAI/dogma --skill transcript-extraction
Clone the repo
git clone --depth 1 https://github.com/EndogenAI/dogma

Made for: Claude Code, Codex.

Per session 95 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 696 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00095 $0.00696
Opus 5 $0.00048 $0.00348
Sonnet 5 $0.00019 $0.00139
Haiku 4.5 $0.00010 $0.00070

Measured yesterday against content hash 96b07a779236, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

transcript-extraction 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 yesterday.

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.

.github/skills/transcript-extraction/SKILL.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

SKILL: Transcript Extraction

This skill governs the extraction and cleaning of transcripts from video sources to support the Endogenous-First axiom (MANIFESTO.md § 1) by ensuring all source signals—even non-textual ones—are distilled into searchable, readable Markdown/text artifacts. It operates under the global constraints of AGENTS.md regarding file-writing and local compute.

Beliefs & Context

  1. Information Density: Video sources often contain unique practitioner insights (demos, conference talks) not found in written docs. Distilling them into text allows for local semantic search and cross-referencing.
  2. LCF Adherence: By caching transcripts in .cache/transcripts/, we avoid repeated network hits and high-token external model reads of video URLs.
  3. Artifact Integrity: Transcripts must be cleaned of redundant timestamps and repetitive filler to preserve context window space.

Workflow & Intentions

1. Identify Video Source

Extract the Video ID from the URL (e.g., J5KTpq7hVn4 from https://youtu.be/J5KTpq7hVn4).

2. Verify Cache First

Check if the transcript already exists in the local cache before attempting extraction.

ls .cache/transcripts/video_<VIDEO_ID>.txt

3. Extraction Protocol

Use the designated internal script to pull the transcript. Prefer manually created English transcripts over auto-generated versions when available.

uv run python scripts/pull_yt_transcript.py <VIDEO_ID> .cache/transcripts/video_<VIDEO_ID>.txt

4. Verification Check

Confirm the file is non-empty and contains readable text.

test -s .cache/transcripts/video_<VIDEO_ID>.txt && head -n 5 .cache/transcripts/video_<VIDEO_ID>.txt

Constraints

  • UTF-8 Only: All extracted transcripts must be stored as UTF-8 encoded plain text.
  • Privacy: Only pull transcripts for public videos.
  • Minimal Metadata: Focus on the spoke word; omit frame data or visual descriptions unless critical to the research question.
  • Pathing: Always store the output in .cache/transcripts/ to ensure it is gitignored but available for the current session.

Read the full file on GitHub · 58 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. yesterday First seen · 58 lines · 0 tokens per session scan A 96b07a779236

Subscribe to this mod's changes

transcript-extraction is a skill published in the GitHub repository EndogenAI/dogma (2 stars, last pushed 8d ago), licensed Apache-2.0. It adds 95 tokens to every session and 696 once invoked, about $0.0005 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.