li-voice
Skill2 starsChanged 44 days agoMITView on GitHub
Build the user's LinkedIn voice profile from three of their own past posts. Use when no voice profile exists yet, when the user says "learn my voice", "write my voice profile", pastes sample posts, or when any other skill asks for a voice profile that is missing. Writes the profile every other skill reads.
In this repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- li-auditSkill
- li-carouselSkill
- li-commentSkill
- li-dmSkill
- li-humanSkill
- li-inboxSkill
- li-planSkill
- li-postSkill
- li-profileSkill
- li-replySkill
- li-repurposeSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Every skill in this pack reads a voice profile before it writes a word.
- Distinctive: line 19
Write to the first of these locations that is writable (prefer the first):
- Distinctive: line 25
If none is writable, print the finished profile in a fenced block and tell the user where to save it.
- Distinctive: line 30
Ask for three of the user's own LinkedIn posts that sound most like them.
- Distinctive: line 31
Not their best-performing posts.
- Very specific: line 32
If they paste fewer than three, work with what you have but say the profile is provisional.
- Distinctive: line 46
Read the three posts and fill in:
- Very specific: line 61
Write the profile using templates/voice.md as the skeleton, with the extracted material filled in.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
li-voice
Every skill in this pack reads a voice profile before it writes a word. This skill creates that file. It is the highest-leverage ten minutes in the whole system: everything downstream is either in the user's voice or it is not.
Where the file lives
Write to the first of these locations that is writable (prefer the first):
~/.agents/linkedin/voice.md(preferred: read by every agent)~/.claude/linkedin/voice.md~/.copilot/linkedin/voice.md
If none is writable, print the finished profile in a fenced block and tell the user where to save it.
The interview
Ask for three of the user's own LinkedIn posts that sound most like them.
Not their best-performing posts; the three that sound most like them talking.
If they paste fewer than three, work with what you have but say the profile
is provisional. If they paste none, do not invent a voice: give them the
blank template from templates/voice.md and stop.
Then ask, in one batched message, whatever the posts do not answer:
- Who are you writing for? (Push for specificity: "agency owners doing $1-5M", not "professionals".)
- What do you sell, if anything?
- Any topics, clients, or claims that are off limits? (NDA, employer policy, regulated industry.)
What to extract from the posts
Read the three posts and fill in:
- Cadence. Sentence length pattern: short and punchy, mixed, long and considered. Quote one representative sentence.
- Diction. Ten words or phrases they actually use; ten they would never use. Be concrete, not "professional tone".
- Devices. What they do repeatedly: numbered lists, one-line paragraphs, rhetorical questions, stories that open in medias res, data-first openings.
- Positions. Two to four things they believe that part of their audience does not. These are where the good posts come from. If the posts do not reveal any, leave this section as questions for the user.
- Emoji, swearing, contractions. State the observed rule plainly.
The file
Write the profile using templates/voice.md as the skeleton, with the
extracted material filled in. End the file with a "Last updated" date and a
note: re-run this skill after every ~20 posts or whenever the user's voice
shifts.
Then confirm in one line what you learned ("Voice profile saved. You write in short punchy paragraphs, open with numbers, and never use emojis.") so the user can correct you before any other skill reads it.
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.