Cosgrove - Speechwriter AI Skill

Write a keynote, eulogy, or toast that sounds like the actual person delivering it, not like a beautifully written speech in someone else's voice.

Cosgrove is a Senior Speechwriter skill built to write for the ear and for one real speaker's actual voice: running an intake interview that captures cadence, vocabulary, and verbal tics before drafting a word, structuring for the specific occasion (a keynote is not a eulogy with different words, a toast is not an investor pitch), building in rhetorical devices made for spoken delivery like rule of three and callback structure, writing short breath-length sentences with deliberate pauses, treating the read-aloud pass as mandatory rather than optional polish, and handing back delivery notes alongside the script.

Who it's for

executives, founders, and individuals preparing a keynote, eulogy, wedding toast, investor pitch, or any address delivered live by a real person.

Key capabilities

• Collects 3-5 real sentences the speaker has actually said and matches rhythm to those instead of drafting in a generic house voice.
• Confirms the structure fits the specific occasion (keynote, eulogy, toast, investor pitch) before writing a single line of the draft.
• Builds in rule of three, anaphora, and callback structure only where they fit the speaker's real voice, never bolted on for effect.
• Keeps sentences short enough to say in one breath and writes at least one clear applause or laugh line per major section.
• Reads every draft aloud, standing, at delivery pace, and rewrites anything that trips the tongue or runs out of breath before calling it done.

How to use it

Paste Cosgrove's SKILL.md into your Claude Project Instructions (or any AI system prompt), then describe the speaker, the occasion, and what the room needs to feel. Works with Claude, ChatGPT, and any AI chat. Under 2 minutes to install.