pitch-package
Skill497 starsChanged 11 months agoMITView on GitHub
The step-back skill for pitching as a person - given an audience (hackathon judges, VCs, investors, customers) and a time slot, it budgets the slot, decides what to build or reuse (deck, demo video, live speech), settles the voice, music, branding and rights questions before anything is built, and orchestrates the pitch-deck, demo-video, and pitch-craft skills into one rehearsed package with a stitch plan, Q&A prep, and a day-of checklist. Use whenever the user says they need to pitch, present, or demo at a hackathon or demo day, to investors or a customer, mentions an N-minute slot, or asks to "prepare the whole pitch" end to end. Not for - a single artifact in isolation (call pitch-deck, demo-video, or pitch-craft directly) or written-only submission forms.
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
STOP: produce no artifacts in this phase.
- Distinctive: line 29
Before writing or building anything, write the budget line:
- Distinctive: line 53
Rehearse against the clock: read the script aloud with a stopwatch once.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Pitch Package - the whole performance
A pitch is one argument delivered through three instruments: the deck (setup), the demo (proof), and the speaker (trust). The demo is the proof; the deck is the setup. This skill's job is sequencing and budget, so the instruments never fight each other. Two rules govern everything: interview before you build, and budget the slot first.
Phase 1 - Interview before you build
STOP: produce no artifacts in this phase. Ask (adapting to what's already known):
- Audience: hackathon judges, VCs, or a customer? What do they explicitly judge or care about (rubric, thesis, their own pain)?
- Slot: total minutes? Is Q&A inside or after it?
- Inventory: what already exists - working product? demo video? deck? scripts? Reuse beats rebuild.
- Demo mode: live product on stage (higher wow, higher risk) or recorded film (deterministic)? Default to recorded unless the live path has been rehearsed with a fallback.
- Speakers: who talks, and are there real credentials to weave into the intro?
- Constraints: venue machine? offline? projector resolution? phone-as-teleprompter?
- Voice and music: a model voice (Kokoro
af_heartat 1.0 is the default) or a human? A music bed, and whose rights? For a model voice, agree to audition one segment two ways after the brief is confirmed, then let them choose; never pick for them. - Branding and rights: the company logo under the picture? A copyright and distribution line, in which words? A disclaimer card for a public copy? Whose logo, if a prospect's brand is in the product? Ask; a brand in every frame cannot be cut out later.
- Where it plays: a portal, a stage, YouTube, a feed? A film with a caption band needs its own 16:9 cut; captions burned in, as a sidecar file, or both.
- Film guardrails: if a demo film will be made or reused, what is it for and what must be burned in - any third-party logo or name in the frame, generic or personalised disclaimer card, a persistent "Demo purposes only" label, the rights line and company mark, and where it may live (private streaming link, our website's unlisted showcase, direct file, public platform)? Optional guardrails, mandatory questions:
demo-video/references/distribution.md.
Reflect the answers back as a one-paragraph brief and get a nod before Phase 2. Skipping this interview is how pitches end up with a beautiful deck for the wrong audience.
Phase 2 - Budget the slot first
Before writing or building anything, write the budget line:
5:00 slot = ~2:30 spoken deck + 2:36 demo video + ~0:10 transitions
(runs ~5:15 as written -> trim points named in the script)The budget decides scope: how many slides get narrated, whether the demo needs a shorter cut, what gets trimmed live. Every downstream artifact inherits this budget. If the pieces don't sum inside the slot on paper, they won't on stage.
Phase 3 - Compose (gap-fill via the other skills)
Pick the audience lens from references/audiences.md - same assets, different lead, emphasis, and close. Then fill only the gaps the inventory revealed:
- Missing demo film -> demo-video skill (storyboard keyed to the proof this audience needs, approved before capture). That skill also owns the motion and annotation layer, so the film can point at what it proves rather than relying on narration alone; budget the extra seconds it costs when you size the demo slot.
- Missing or stale deck -> pitch-deck skill (demo-slot slide included; appendix slides for this audience's hard questions).
- The live speech -> pitch-craft skill, LIVE register: hook first, team intro after the hook, the demo-break marker stating what the video covers, delivery cues, named trim points.
One throughline threads all three artifacts - if the deck's tagline, the video's close, and the speech's refrain differ, unify them before rehearsal.
The joins are part of the throughline work. Wherever the film or the talk changes person, case or topic, a transition line says so before the screen does, and nowhere else (pitch-craft's references/storytelling.md). The close lands the throughline, then the ask for customers or investors, or the name for judges, as the last thing heard; for a film that is a recap card walked by the voice, with the pilot line last (demo-video's references/structure.md).
Phase 4 - Rehearse and ship
- Stitch plan per
references/stitching.md(which variant: fully recorded film, live-with-recorded-demo, fully live) plus the fallback ladder. - Rehearse against the clock: read the script aloud with a stopwatch once; apply the named trims if over. Adjust the script, never the speaking pace.
- Q&A prep: appendix slides + a one-page honest-defense brief (the questions you fear, answered with what's true - overclaims die here, before a judge finds them).
- Day-of checklist: local copies of deck AND video on the presenting machine (no network dependency), test in an incognito window at venue resolution, backup speaker for each section, the teleprompter (phone-friendly script page) charged and loaded.
- Deliverables: the film, its 16:9 cut if it carries a band, the
.srt, the timed script, the thumbnail, and one line naming the voice, the speed, the music track and its licence (demo-video'sreferences/soundtrack.md).
References
references/audiences.md- the three lenses (judges / VC / customer): what each judges, what leads, what closes.references/stitching.md- assembly variants, ffmpeg stitch recipe, fallback ladder, phone-teleprompter pattern.
In this repository17 other files it gives its agents.
- agentic-system-designSkill
- cross-app-shared-authSkill
- demo-videoSkill
- excalidrawSkill
- google-analytics-setupSkill
- htmldropSkill
- landing-page-gtmSkill
- meta-pixel-setupSkill
- motion-videoSkill
- pitch-craftSkill
- pitch-deckSkill
- project-showcaseSkill
- scribeSkill
- secure-by-designSkill
- self-improving-systemsSkill
- streakSkill
- uat-testingSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

