- Home
- Skills
- Productivity
- deep-read
Works with the AI Agents you already use
Deep Read
Hand an agent a 3-hour transcript and it usually reads part of it, then gives you a summary that *looks* complete.
Free
deep-read
Example session with this skill installed
/deep-read ~/Downloads/how-to-build-a-personal-brand-with-ai.srt — it's a 3-hour course. Read all of it, don't skip anything, and tell me what's actually worth taking away.
- Read your context and instructions
- Compiled the deep-read
Notes: "How to Build a Personal Brand with AI (Full Course)"
Coverage: read transcript.txt in full. Folded to 200-character lines (1,309 lines), read in 4 chunks: 1–330, 330–660, 660–990, 990–1309. Nothing skipped.
Source: 5,550 SRT cues merged into timestamped paragraphs, 2 h 53 min, ~41k words. Numbers are as spoken by the author and not independently verified; ⚠️ marks contradictions or unclear passages.
What it is saying
Products are getting cheap to build with AI, so distribution is the scarce asset, and a personal account beats a company account at getting it. The course is the author's playbook for producing enough content to get through the "trust period" (roughly 100 posts before reach arrives), using Claude skills to keep up four Instagram posts a day.
Worth taking away
- The real substance is in three places: YouTube's three metrics and packaging plus section-by-section hooks; ManyChat / GoHighLevel automation setup; and Meta ads setup with unit economics.
- The skill method transfers directly: run a task end to end, iterate until it's right, then have the agent read the whole session back and save it as a skill.
- The course is itself a funnel (community, agency, affiliate links), which is exactly the "one long course a month" play it teaches.
Flagged while reading
| Where | Issue |
|---|---|
| [Meta ads] | Transcript says cost per sale "$641"; later "about $6 per trial" means it is $6.41. |
| [ManyChat] | Welcome DM says the community has "165" members; later it is 165,000. Caption error or a different group. |
| [Intro] | Promises repurposing to X / LinkedIn / Threads; never covered. |
| [Meta ads] | "I'll explain why you need more ads than the cold campaign" — never explained. |
| [ManyChat] | Says he isn't sponsored, but the link is an affiliate link (trial extended 14 → 30 days). |
Caption fixes applied while reading
claw / cla / plaud → Claude · Hickfield / Hixville → Higgsfield · Minihat / Many Shadow → ManyChat · school → Skool · rorowass → ROAS
Next steps
- Section-by-section notes (17 sections with timestamps) follow in
notes.md. - Ask follow-ups against the notes; anything the notes don't cover is checked in the source, not recalled.
Connects securely to your tools. The creator never sees your data.
About this skill
Hand an agent a 3-hour transcript and it usually reads part of it, then gives you a summary that looks complete. A single read errors above ~25k tokens, Read returns only the first 2000 lines, a paragraph-per-line transcript can't be split by line, or the agent just greps and summarizes.
deep-read fixes the workflow:
- A bundled script (
prepare.py, Python standard library only) turns the source into a reading copy with short lines and a chunk plan that fits the read limit, estimated per language (CJK text costs ~4x the tokens of English per character). - The agent reads every chunk in order itself: no subagent summaries, no sampling.
- Notes are written after each chunk, so progress survives context compaction.
- Once every line is read, the notes open with what it is saying: the main point, how the parts build it, where it is strong or weak, and what is worth taking away. Fact-checking depth follows the kind of source: a speech is read for its meaning, a market call for its evidence.
- A second agent reads the source from scratch without seeing the notes; every disagreement is checked against the source.
The notes end with a coverage line: file, line count, and the line range of each chunk.
Works on: course, podcast and speech transcripts, SRT/VTT subtitles (become timestamped paragraphs), long articles, meeting notes, PDFs (text layer via pdftotext; scanned PDFs page by page).
How it differs: "full output" skills stop agents from writing less than asked; deep-read stops them from reading less than given. Analysis-template skills apply a framework; deep-read makes reading every line, with proof, the floor, and asks for understanding in the source's own terms on top of it.
Open source on GitHub: https://github.com/max1874/skills
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 8 days ago
- Free with an Agensi account
Requires