Twelve skills for conference speakers covering rhetoric analysis, talk authoring, deck building and export, CFP submissions, publishing, illustrations, shownotes, screencast production, and vertical shorts, informed by a 113-entry Presentation Patterns taxonomy (83 observable: 64 patterns + 19 antipatterns; 30 unobservable: 21 patterns + 9 antipatterns).
92%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Medium
Suggest reviewing before use
A curated list of failure modes specific to the shownotes Jekyll pipeline. Each entry: the mistake, what visually happens, and the right way.
Don't: Commit an exported PDF under assets/slides/ and point the Slides
field at that site URL. The embed include recognizes provider URLs; a static
PDF URL falls back to a link even when the PDF download returns HTTP 200.
Do: Follow Slide PDF Hosting. Keep the PDF in the talk's Drive folder and verify its checksum matches the local export. Confirm public reader access and publish the Drive preview link. Inspect the built iframe and the live rendering. Do not add native-PDF hosting to the site to work around an incorrect publication destination.
Don't:
**Video:** TBD
**Video:** [Coming soon](#)
**Video:** [TODO]
**Video:** Recording coming soonWhat happens: Any non-empty extracted_video value fires the
"Video Available" badge. The "Video Coming Soon" badge that the
speaker actually wants never renders. The video-embed section then
tries to embed "TBD" as a URL, producing a broken iframe.
Do: Omit the **Video:** line entirely. When the video lands,
add the line. The badge automatically flips on rebuild.
Don't:
**Slides:** [View Slides](#)
**Slides:** [View Slides](#) <!-- TODO -->
**Slides:** TBDWhat happens: Identical to the Video gotcha — extracted_slides
becomes # (truthy), the layout's media section renders, and the
slides embed tries to load # as a URL. The page shows a broken
iframe inside the "Slides" panel.
Do: Omit the **Slides:** line entirely until you have a real
URL. The talk page is fully valid without a slides section. When
the URL lands, add the line — no other change needed.
thumbnail_url: in FrontmatterDon't:
---
layout: talk
thumbnail_url: /assets/images/thumbnails/my-talk-thumbnail.png
---What happens: Nothing useful. Thumbnails are resolved entirely
by the filename-slug naming convention (/assets/images/thumbnails/{slug}-thumbnail.png)
in index.md, _layouts/default.html, and
_includes/embedded_resource.html. The frontmatter field is checked
only as a truthiness signal for the homepage's "Highlighted
Presentations" inclusion filter (index.md:104), and that's
redundant with extracted_slides/extracted_video — both of which
are already populated from the body field block on any normal talk.
The field misleads a reader into thinking they can override the thumbnail path. They can't. Setting it to a non-existent path doesn't make a thumbnail appear; setting it to a different path doesn't change which file the templates load.
Do: Skip the frontmatter field. Drop the thumbnail file at the
conventional path ({slug}-thumbnail.png). The templates find it
automatically; missing thumbnails fall back via onerror handler
to /assets/images/placeholder-thumbnail.svg.
Don't:
## Abstract
This is the first paragraph of the abstract.
This is the second paragraph.
### Sub-section in Abstract
This is a third paragraph under a sub-heading.
- Bullet one
- Bullet twoWhat happens: Everything between ## Abstract and the next ##
heading gets joined into ONE STRING with whitespace collapsed.
Sub-headings and lists become run-on text:
"This is the first paragraph of the abstract. This is the second paragraph. ### Sub-section in Abstract This is a third paragraph under a sub-heading. - Bullet one - Bullet two"
After markdownify, the ### and - characters mostly survive as
literal text (markdownify on a single line doesn't reconstruct
headings or lists).
Do: One paragraph of flowing prose. If the talk has multiple movements to describe, write a single dense paragraph that walks them in sequence. Save the structured outline for the talk itself; the abstract is the elevator pitch.
Don't:
**Slides:** <google-drive-url>
**Video:** <youtube-url>What happens: The parser's URL extraction regex
\[([^\]]+)\]\(([^)]+)\) doesn't match — it expects the
[text](url) form. The full plain-text value ends up in
extracted_slides / extracted_video. The status badge fires
(string is truthy), but the embed include then tries to detect URL
patterns inside the plain string. It often "works" for YouTube — the
youtu.be/ or youtube.com/watch?v= substrings still trigger the
embed branch. For Google Drive PDFs the pattern match is less
forgiving.
Do: Always wrap in markdown link syntax:
**Slides:** [View Slides](<google-drive-url>)
**Video:** [View Video](<youtube-url>)Don't:
# Talk Title
…field block…
# Section One
Body for section one…
# Section Two
Body for section two…What happens: Only the first # line becomes
extracted_title. The subsequent H1s render as H1s in the body —
the layout's CSS sizes them as H1, which looks visually odd
(two title-sized headings on one talk page).
Do: One H1 (the talk title). Use ## for second-level
headings — but only ## Abstract and ## Resources are
parser-aware sections. Other ## Headings render but aren't
extracted; they appear as body content between the field block and
the abstract OR after the Resources section.
Don't:
---
layout: talk
title: My Talk Title
---
# My Talk Title
…What happens: The layout uses extracted_title as the visible
H1. The frontmatter title: overrides only the page title in some
rendering paths. The body H1 still renders. End result: title
appears once at the top (from the layout's H1 emission of
extracted_title) but the body # My Talk Title also renders below
the field block, producing duplication.
Worse: if you omit the body H1 thinking the frontmatter title
suffices, the parser's extract_title_from_content falls through to
"Untitled Talk" — and the LAYOUT's display logic compares
extracted_title against page.title to decide whether to humanize
the slug. The conditionals don't agree, and the page is broken in
subtle ways.
Do: Put the title in a # Title H1 at the top of the body. Skip
the frontmatter title field.
Don't:
## Resources
- [link](https://example.org)
## Abstract
Prose.What happens: The parser collects abstract from ## Abstract
until the next ## heading. If Resources comes BEFORE Abstract,
Resources isn't reached by the abstract scanner (it stops at the
boundary, which here is end-of-file or another section). But the
Resources scanner finds ## Resources and captures everything
after it — including the ## Abstract line and its body. So:
extracted_abstract = "Prose." ✓ (this works)extracted_resources = "- link ## Abstract Prose." ✗
(Abstract content is folded into Resources)Do: Order sections Abstract first, Resources last. Parser behavior is sequential — Resources captures to end-of-file.
Don't:
# Talk Title
**Conference:** Foo 2026
**Date:** 2026-01-01
This is a personal note I want at the top of the page.
A presentation at Foo 2026 by {{ site.speaker.name }}
## Abstract
Prose.What happens: The "personal note" line is captured by the
abstract's legacy fallback IF there's no ## Abstract section, OR
ignored otherwise. With ## Abstract present, this note line just
renders as body text between the field block and the presentation
context — invisible markdown structure but it does show up in the
final HTML.
The presentation context parser specifically looks for "A presentation at" as the start marker — so the note above is fine visually if you want it. Just be aware it's not "metadata"; it's body text rendered as-is.
Do: If you want a pre-abstract callout, use it deliberately and preview the rendered page. Most authors skip this and go field-block → presentation-context → abstract → resources.
Don't expect:
**Subtitle:** Engineering Context for Agentic AIWhat happens: The parser doesn't extract Subtitle — it's not in the metadata-field whitelist. The layout doesn't render a subtitle either. The line shows up as inline body text after the field block, which looks awkward.
Do: If a subtitle matters for the talk identity, embed it in the
H1 (# Talk Title — Subtitle Here) or in the first sentence of the
abstract. Don't author a Subtitle field expecting it to render
separately.
Same as Subtitle. There's no parser support for **Co-speaker:** —
it renders as plain body text. The presentation context paragraph is
where co-speakers go: "by {{ site.speaker.name }} and Co-Name".
Don't: Open the existing file, read it, then write a fresh composition with the "improved" version.
What happens: Speakers regularly hand-edit shownotes after publish — typo fixes, resource additions, link updates. A re-author wipes those edits silently. The talk page reverts to whatever the generator's mental model was at write-time, which may be days behind.
Do: Read the existing file. Modify only the lines that need to
change (typically: adding the **Video:** line when the recording
lands, appending a new resource link). Preserve everything else.
If a major rewrite IS warranted, explicitly tell the speaker the existing file will be overwritten and confirm before proceeding.
<!-- TODO --> Markers in Published ContentDon't:
<!-- TODO: confirm slide/video URLs + generate thumbnail before publish -->
# Talk Title
**Slides:** [View Slides](#) <!-- TODO -->
**Video:** [Watch Video](#) <!-- TODO -->What happens:
<!-- TODO --> after the markdown link is captured
by the parser's value group (^\*\*Slides:\*\*\s*(.+)$ matches
to end-of-line) — extracted_slides ends up containing the
comment text in addition to the link. The URL extraction regex
still finds # in the bracket-paren form, but the wider value
pollutes debug output# placeholders intactDo: If the work is incomplete, omit the incomplete fields. The talk page renders fine without Slides or Video sections. When the URLs land, add the lines with real values — no comment markers.
For tracking incomplete work, use a checklist in a PR description
or an issue, not a comment in the committed file. Speakers can grep
the repo for TODO if needed, but committed <!-- TODO --> in
published files inevitably stays there.
.tessl-plugin
rules
skills
cfp-writer
references
illustrations
presentation-builder
presentation-creator
references
patterns
build
deliver
prepare
scripts
presentation-publisher
screencast-editor
screencast-recorder
shorts-producer
shownotes-publisher
vault-clarification
vault-ingress
references
scripts
vault-profile