- Home
- Skills
- Research & Analysis
- Slide Architect: Presentation-Ready Research Talks
More screenshots
Works with the AI Agents you already use
Slide Architect: Presentation-Ready Research Talks
Paper figures shrunk onto slides, headlines like "Results" and a Methods section before any result lose them.
$12
Slide Architect: Presentation-Ready Research Talks
Example session with this skill installed
I'm presenting this paper (paper.md) as a 12-minute talk at the Regional Air Quality Workshop next month: the audience is a mix of atmospheric scientists and city environment staff, with 3 minutes of questions after the 12. Can you build the slides from the paper and my figures in figures/? The data and the script that made the figures are in data/ and scripts/.
- Read your context and instructions
- Compiled the slide architect
- Generated the data export
I've built the slides: 12 main slides plus 2 backups. The build reports no overflow or blurry figures, the content check has nothing left to fix, and all 86 numbers on the slides and in the notes trace back to `paper.md` or `data/daily.csv`. I also recomputed every number in the paper from the data, and they all match. **Files** (in `talk/`): `deck.pptx` to edit and present from, with speaker notes on every slide; `render/deck.pdf` as a backup copy; `deck_notes.md`; `make_slide_figures.py`, which redraws your three figures from the data at slide size (same correction, κ = 0.35, same held-out days); `sources.md` and `storyboard.md`. **Main message (slide 3, repeated on the end slide):** a one-parameter humidity correction cuts sensor error by 69% on held-out days (RMSE 16.8 → 5.2 µg/m³). **Timing:** 10.3 of 12 minutes planned. Problem and finding 2.1 min, evidence 6.4 min, limits 1 min, take-home 0.8 min. **Things to know before you present** - **Your abstract has a wrong number.** It says the correction cuts error "by three-quarters", but your own results and data give 69%. The slides use 69%. - **There is no contact email on the end slide**, because none was in the material. - **The end slide gives advice the study doesn't test**, so I worded it as "45 days here … test it at your own site". **Fresh-reader test:** a separate agent saw only the slide images, stated the main message correctly and read every slide as intended. From its notes I fixed the title's line breaks, one headline and a clipped peak. **Likely question:** does one κ suit every sensor? In your data it doesn't (one sensor's error rises after correction). That isn't in the paper, so it stays off the slides; the details are in `sources.md`.
daily.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
## What it does A research talk is an argument made out loud to people who cannot turn back a page. Paper figures shrunk onto slides, headlines like "Results" and a Methods section before any result lose them. This skill turns your agent into a careful talk designer. Give it your paper or abstract, your figures, and (if you have them) the data and plotting script, and it: - **Builds the talk around your finding**: the problem in the audience's terms, the answer stated by slide 3 or 4 with its number, evidence blocks that earn it, a limits slide, and a take-home that stays up during questions. Sized for the setting (lightning talk, conference talk, seminar, job talk, defence, lab meeting) with a time plan that fits the slot. - **Writes every headline as a claim**: a full sentence stating what the slide shows ("Retention falls from 80% in small storms to 53% in large ones"), backed by one figure, one number or a short list. - **Keeps every number traceable**: an inventory of your results with their locations, and a check of every number on the slides and in the notes against your paper and data (found, rounded or a declared calculation). Nothing is invented; inconsistencies in the material are reported to you. - **Redraws your figures for the screen**: from a copy of your script and your data, at the exact size each figure occupies on the slide, with 20 pt text, the data the headline is about in one accent colour, and direct labels instead of legends. Your originals are never touched. - **Builds an editable .pptx** with speaker notes (PowerPoint, Keynote, LibreOffice, Google Slides), plus an HTML preview, a PDF and an image of every slide. - **Measures the layout** in a real browser engine: text that runs past its box, headlines over two lines, figures that will look soft on a projector, shrink below 18 pt or leave their space half empty. - **Looks at every slide**, then runs a fresh-reader test: a subagent sees only the slide images and says what the talk claims. It can also review an existing .pptx: topic titles, wordy slides, small text, legends, pies and 3D charts, with a proposed restructure. ## What's included - `SKILL.md`: the workflow (facts, evidence inventory, finding and arc, headline-first storyboard, figures at slide size, build and checks, look at every slide, fresh-reader test, hand-over), integrity rules and slide design rules - `scripts/build_deck.py`: spec to .pptx, HTML, notes, PDF, slide images and a measured layout report. Finds Chrome, Chromium, Edge or Brave on macOS, Linux and Windows. - `scripts/trace_numbers.py`: checks every number on the slides and in the notes against your material - `scripts/deck_lint.py`: content rules for a spec or an existing .pptx, including a tested claim-or-topic check for headlines - a slide style block (in SKILL.md): matplotlib settings for figures at slide size, and equations rendered to images. The figure script you keep carries it, so it redraws your figures with matplotlib and numpy alone - `references/talk-arcs.md` (structure, sizes and time shares by setting) and `references/spec.md` (every spec field and report line) - `examples/`: a complete synthetic ten-minute talk (paper, data, figure script, spec) ## Part of the Aurora Lab research pipeline Works entirely on its own. With a `research/` folder it reads figures, results, claims and the paper from there and writes to `research/talk/`. Pairs well with Aurora Lab's Talk Coach, Q&A Prep, Poster Layout and Paper Figures, but needs none of them.
It builds talks from material you already have; it does not run new analyses, and it cannot redraw a figure without its data (an image-only figure is used as is if it is legible, or you are asked for the data or a larger export). The deck uses one built-in style: white slides, one sans-serif font, one accent colour, tinted section slides; it does not import institutional PowerPoint templates or produce Beamer/LaTeX, and animations are not generated. The .pptx is fully editable, but its text is laid out to match the measured HTML preview, so PowerPoint, Keynote or LibreOffice may wrap lines slightly differently: leave a little room. Rendering needs Chrome, Chromium, Edge or Brave; Firefox and Safari cannot be driven, and without one there are no slide images or text checks. The number check confirms that each number appears in your material (or is a declared calculation), not that it is used with the right meaning, and numbers inside images or written as words are checked by the agent, not the script. The headline check is a heuristic for English (with basic Spanish) and only warns. Talk lengths and formats differ between venues and institutions: check your event's current instructions.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 27 days ago
Requires
Permissions
File Scopes
Reads the paper, figures, data and plotting scripts you point it at (with a research/ folder: research/paper/, research/figures/, research/results/, research/claims.md). Writes the sources list, storyboard, spec, redrawn figures, a copy of your plotting script, the .pptx, HTML, notes, PDF and slide images to talk/ (or research/talk/). Runs a Chrome-family browser headless, locally, with a temporary profile, only to print the PDF, render the slides and measure the layout. Never modifies your paper, figures, data or scripts. No network access; installs nothing without asking, and only into the project's environment.