find-my-voice
Skill2 starsChanged 37 days agoMITView on GitHub
Capture the user's writing voice for their portfolio by extracting it from things they actually wrote, or a short interview if they have nothing. Use for "find my voice" (often with posts or samples they wrote), "what's my voice", "make my portfolio sound like me", setting up a voice profile, or before writing portfolio content. Not for general editing or matching a style outside the portfolio.
In this repository11 other files it gives its agents.
- anonymizerSkill
- content-strategistSkill
- portfolio-copywriterSkill
- quality-checkSkill
- refreshSkill
- setupSkill
- skill-improverSkill
- story-adapterSkill
- voice-guardianSkill
- web-content-optimizerSkill
- website-expertSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Extract, don't ask: If you ask someone "what's your voice?", they answer in generic LinkedIn-speak ("clear, strategic, a little warm") — …
- Distinctive: line 24
If they truly have nothing written, use the fallback below.
- Distinctive: line 28
Read the samples and extract concrete, observable patterns — not adjectives:
- Very specific: line 49
Write a detailed ## Voice Profile into CLAUDE.md — not one word, a usable spec.
- Distinctive: line 80
Use the archetypes below only as a last-resort nudge if they freeze, never as the answer:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Find My Voice
Capture a rich, specific voice profile by analyzing what the user has actually written. A real voice cannot be picked from a list of five archetypes — it has to be observed. This profile becomes the target every other skill writes toward.
Core principle
Extract, don't ask. If you ask someone "what's your voice?", they answer in generic LinkedIn-speak ("clear, strategic, a little warm") — the exact thing this kit exists to kill. People can't describe their voice; they can only demonstrate it. So work from samples.
Process
1. Get samples (the default path)
If CLAUDE.md has no # Portfolio Profile, run setup first. Then read CLAUDE.md → Source Material. If writing samples are saved (e.g. source/writing-samples.md), use them and ask for more only if there are fewer than 2 — and never if the user already said that's all they have. If it says "none", go straight to the fallback (unless they just gave you samples).
Otherwise ask for 2–3 things the user has written: LinkedIn posts, a newsletter, a talk transcript, a blog, even a few real emails. Read whatever they give (file, paste, or link), save pasted samples to source/writing-samples.md, and record that path in Source Material.
If they truly have nothing written, use the fallback below. With only one short sample, also run the fallback interview.
2. Analyze for a specific fingerprint
Read the samples and extract concrete, observable patterns — not adjectives:
- Openings — how do they start? (hook, question, blunt statement, story)
- Sentence rhythm — short and punchy? long and winding? mixed?
- Signature vocabulary — words and phrases they reach for repeatedly
- Tone markers — where they go warm, dry, funny, direct, contrarian
- Structure habits — lists, one-liners, build-ups, how they land a point
- What they avoid — jargon? adjectives? hedging? exclamation marks?
3. Reflect it back, specifically
Show the user their own fingerprint with evidence, e.g.:
"Here's how you actually write: you open with a contrarian one-liner, keep sentences short, lean on 'the real story is' and 'basically', and you almost never use adjectives. Dry, confident, a little blunt. Sound right?"
Quote real patterns from their samples. Let them confirm or nudge.
4. Adjust for register
Name the one real gap: their casual voice (newsletter, posts) may not be their portfolio voice. Ask if they want it tuned up a notch (more polished) or kept exactly as-is. Capture the raw voice, then note the target register.
5. Store a rich profile
Write a detailed ## Voice Profile into CLAUDE.md — not one word, a usable spec. Keep client names, codenames, people, and figures out of it: pick sample lines and signature phrases that show the voice without them.
## Voice Profile
**One-line:** [e.g. "Dry, blunt, confident — short sentences, no fluff"]
**Register:** [casual / polished-casual / formal]
**Opens with:** [pattern]
**Sentence style:** [pattern]
**Signature phrases:** [list]
**Leans into:** [warm / dry / contrarian / etc.]
**Avoids:** [jargon / adjectives / hedging / etc.]
**Sample lines (real):**
- "[a real sentence they wrote]"
- "[another]"This rich profile is what story-adapter, refresh, and any content step check against — and what keeps the output from sounding like everyone else.
Fallback (no writing samples)
First run a short voice interview — extract from how they speak, since spoken answers reveal voice just as well. Ask 3–4 questions and let them talk freely:
- "Tell me about a project you're proud of — what was broken, and what did you do?"
- "How would you explain what you do to someone at a dinner party?"
- "What's a take in your field you'd defend when others push back?"
- "What kind of writing makes you cringe?" (reveals what they avoid)
Extract the same specific fingerprint from their answers — openings, rhythm, signature phrases, tone, what they avoid — reflect it back, and write the rich profile.
Save their answers to source/voice-interview.md and list it under Source Material → Notes, so story-adapter can reuse any story they told. Anything they say is private or under NDA goes under Privacy → Never name.
Use the archetypes below only as a last-resort nudge if they freeze, never as the answer:
- Builder — "I noticed X was broken, so I built Y"
- Strategist — "I saw the gap between X and Y, so I reframed it"
- Operator — "The process was a mess; I redesigned it and we shipped"
- Narrator / Marketer — "Here's why this matters and who it's for"
- Technical — "The system couldn't handle X, so I changed the architecture"
Even if you use an archetype as a nudge, still extract the real fingerprint from how they talk — never leave them as just a label.
Handoff
"Voice captured. From here, everything I write — case studies, updates — gets checked against this so it stays unmistakably you. Next: story-adapter to turn your work into case studies."
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.

