YouTube Content Skill for Hermes Agent: Transcripts to Summaries

The YouTube Content skill turns a video link into something you can read, quote, or publish. Hand Hermes Agent a YouTube URL and it can pull the transcript, then reshape it into chapters, a summary, a thread, a blog post, or a set of timestamped quotes. It is the skill to reach for when the source material lives in a video and you need text out of it without watching.

What it does

The skill wraps a helper script that fetches a video's transcript and returns it in a form you can process. It accepts the usual YouTube URL shapes, including short links, shorts, embeds, live links, and a bare 11-character video ID. Once the transcript is in hand, the skill formats it according to what the user asked for: a chapter list, a short overview, per-chapter summaries, a numbered thread, a full article, or notable quotes with timestamps. If no format is specified, it defaults to a summary.

Before you start

The skill ships bundled with Hermes Agent, so there is nothing to install from a marketplace. What you do need is a Python environment prepared through the project's package manager. The source is explicit that you should use the Python from a PM-prepared Hermes source checkout, and that the youtube extra declares the helper's dependency. Do not install packages into Hermes with raw pip or with a project-discovering uv run.

From that checkout, follow the isolated development-home setup in Package Management first, then prepare the extra and reactivate before running the helper:

source ./activate
python -c "import pm; pm.sync_venv(['youtube'], explicit=True)"
source ./activate
python -c "import youtube_transcript_api; print(youtube_transcript_api.__file__)"

On Windows, use . .\activate.ps1 instead of source ./activate. If the terminal runs on a different host or in a sandbox, use an explicitly isolated helper environment there, not the agent's production environment. Run every command below with the interpreter whose import check succeeded.

Fetching a transcript

SKILL_DIR is the directory containing this SKILL.md file. The helper script lives under scripts/ inside it. The four invocations below cover the common cases: structured JSON with metadata, plain text for piping, timestamped output, and a language preference with fallback.

# JSON output with metadata
python SKILL_DIR/scripts/fetch_transcript.py "https://youtube.com/watch?v=VIDEO_ID"

# Plain text (good for piping into further processing)
python SKILL_DIR/scripts/fetch_transcript.py "URL" --text-only

# With timestamps
python SKILL_DIR/scripts/fetch_transcript.py "URL" --timestamps

# Specific language with fallback chain
python SKILL_DIR/scripts/fetch_transcript.py "URL" --language tr,en

Use --text-only when the next step is another script or a summarization pass that does not need structure. Use --timestamps when you plan to produce chapters or quotes, since those formats depend on knowing where in the video a line occurs. The --language flag takes a comma-separated chain, so tr,en asks for Turkish first and falls back to English.

Output formats

After fetching the transcript, format it based on what the user asks for:

• Chapters: Group by topic shifts, output timestamped chapter list
• Summary: Concise 5-10 sentence overview of the entire video
• Chapter summaries: Chapters with a short paragraph summary for each
• Thread: Twitter/X thread format, numbered posts, each under 280 chars
• Blog post: Full article with title, sections, and key takeaways
• Quotes: Notable quotes with timestamps

Example, Chapters Output

00:00 Introduction — host opens with the problem statement
03:45 Background — prior work and why existing solutions fall short
12:20 Core method — walkthrough of the proposed approach
24:10 Results — benchmark comparisons and key takeaways
31:55 Q&A — audience questions on scalability and next steps

Workflow

1. Fetch the transcript using terminal and the prepared Python with --text-only --timestamps.
2. Validate: confirm the output is non-empty and in the expected language. If empty, retry without --language to get any available transcript. If still empty, tell the user the video likely has transcripts disabled.
3. Chunk if needed: if the transcript exceeds ~50K characters, split into overlapping chunks (~40K with 2K overlap) and summarize each chunk before merging.
4. Transform into the requested output format. If the user did not specify a format, default to a summary.
5. Verify: re-read the transformed output to check for coherence, correct timestamps, and completeness before presenting.

The chunking step matters for long videos. A single pass over a very long transcript tends to lose detail near the end, so the skill splits it into overlapping windows and merges the summaries. The overlap keeps context from being cut mid-thought at a boundary.

When not to use it

This skill is for YouTube video content. If the source is a page, a PDF, or an audio file that is not on YouTube, it is the wrong tool. If the video has no transcript available, the skill cannot manufacture one; the error handling below covers what to do instead.

Limits and gotchas

The source lists the failure modes you are likely to hit:

• Transcript disabled: tell the user; suggest they check if subtitles are available on the video page.
• Private/unavailable video: relay the error and ask the user to verify the URL.
• No matching language: retry without --language to fetch any available transcript, then note the actual language to the user.
• Dependency missing: repeat PM preparation and reactivation above, then verify the helper uses that Python. Do not repair the selected generation with pip.

The dependency rule is the one people trip over. Installing youtube_transcript_api with pip into the wrong environment will appear to work until the helper runs under a different interpreter. The import check in the setup block is there to confirm you are on the right one before you spend time debugging a fetch.

What pairs with it

This skill sits in the Media category alongside the other bundled media skills. For the environment side, the Package Management reference covers the isolated development-home setup the setup block depends on. If you are building a publishing workflow, pair the transcript fetch with whatever writing or scheduling skill you already use; the skill's job ends at producing the formatted text.