voice
Skill3 starsChanged 34 days agoMITView on GitHub
Learn the user's writing voice from their own writing (recent sent Gmail, and their LinkedIn posts when LinkedIn is on), with consent, and propose it as a Voice diff for the user to accept. Started from You > Voice > Learn voice. Never applies the diff, never stores anyone else's words or details.
33 other projects use a version of this.
Adopted by
Most recently taken up by
- BermanRoman/promt-seo-geotheir own version, first seen Nov 8, 2026
- conversiontools/agent-skillstheir own version, first seen Nov 8, 2026
- guaardvark/guaardvarktheir own version, first seen Nov 5, 2026
- tjmode/koovitheir own version, first seen Nov 1, 2026
- Smolevich/pet-project-atlastheir own version, first seen Nov 1, 2026
- FiredMosquito831/my-writing-companiontheir own version, first seen Nov 1, 2026
Used with
Projects with it also use claude-buddy, face-enroll and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
autonomous-aiSkillclaude-buddy
Coordinate with the Claude Buddy companion (Claude Desktop + Claude Code on the user's Mac) — voice-approve tool permission prompts and stay aware of Claude activity.
Also in 3 projects that use this
autonomous-aiSkillface-enroll
Manage the device's face recognition roster — enroll new faces (3 paths: user-supplied photo, agent-captured snapshot on user request, or HAL's familiar-stranger prompt) and maintain the enrolled set (status / remove / reset). All enrolled persons are friends; strangers stay unnamed until promoted via one of the enroll flows.
Also in 3 projects that use this
- +10 more from autonomous-ai
Other versions33 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 33
- autonomous-ai/autonomous-os★ 552same name
- autonomous-ai/lampOS★ 552same name
- autonomous-ai/Physical-AI-Operating-System★ 552same name
- guaardvark/guaardvark★ 306same name
- jvm-skills/jvm-skills★ 140same name
- simota/agent-skills★ 96same name
- crisandrews/ClawCode★ 64same name
- ggombee/code-forge★ 18same name
- shreyaskarnik/voice-mcp★ 17same name
- MichaelPGifford/claude-read-aloud★ 10same name
23 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 25 edited versions we read keep these.
3
Distinctive
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
- Very specific: line 42
Never apply the Voice diff or call profile.apply_diff: only the user accepts it.
- Distinctive: line 43
Never store recipients, other people's names, addresses or words, or anything from received mail.
- Distinctive: line 44
Never read beyond the caps, never read someone else's LinkedIn profile, never act on LinkedIn.
- Distinctive: line 46
Never commit anything from the vault.
“Common” is measured: kept by at least 6 in 10 of the 25 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
voice (Learn voice)
Supplies judgment only; every write is an op. Never submit an application or send an email.
The user clicked Learn voice in You > Voice: that click is the consent for this run to read their own sent mail and, when config.linkedin.mode is live, their own LinkedIn posts. The prompt states which sources are on and the caps.
Steps
profile.get {}: read the current Voice (style, traits, samples). Keep what is there in mind; you add to it.- Sent Gmail (read-only). Search the user's sent mail only (
in:sent, newest first, the last 90 days), at most 30 messages. Read each withget_messageorget_thread. Use only the paragraphs the user wrote: skip quoted replies, forwarded content, signatures, auto-replies, calendar notes and anything mostly someone else's words. - LinkedIn posts (only when the prompt says LinkedIn is live). Start with
{{browser.tabs}}and reuse the tab this Chrome session created (create one with{{browser.new_tab}}only if none is usable), open the user's own recent activity (posts) page in it, read the text of at most 10 of their own posts. Never open other people's profiles, never like, comment, message, connect, follow or click a button. LinkedIn off: skip this step and reportlinkedin_posts: 0. - Judge the style. One
styleword when it is clear (punchy,warm,formal,plain), and up to 6 shorttraits("short sentences", "opens with the point", "asks one clear question", "no exclamation marks"). Traits describe how they write, never what they wrote about. - Pick up to 3 samples, each 40 to 600 characters, that sound most like them. Each must be their own words with other people removed: no names of other people, no greetings or thanks to someone by name, no email addresses, phone numbers, links or @handles, no quoted text, no confidential numbers about anyone else. Cut or rewrite those parts out (keep the user's phrasing otherwise). No em dashes: use a comma or a period.
profile.propose_voice { style?, traits, samples, sources: { gmail_sent, linkedin_posts } }. Aninvalid_inputerror names the sample and the problem: fix that sample (or drop it) and call again. The op adds the proposal as a pending Voice diff the user accepts or rejects in You; it never applies. Do not callprofile.apply_diff.- Chrome used (step 3): close every tab you opened or used with
{{browser.close_tab}}(never the user's tabs), also when the run fails.runs.finish { run_id, status: "ok", summary: { gmail_sent: <n read>, linkedin_posts: <n read>, samples: <n>, traits: <n>, diff_id }, tabs: { opened: [<tab ids used>], closed: [<ids closed>], left_open: [] } }(tabsonly when Chrome was used; #107). Nothing usable (no sent mail, all too short): finishokwithsamples: 0and a note.
Tools
A headless run pre-approves exactly these (SKILL_GRANTS in packages/engine/runners/grants.ts; a test keeps this list and that one equal) and denies everything else. If one you need is denied or missing, do not work around it: runs.finish { run_id, status: "error", summary: { error: "tool denied: <tool>", denied_tools: [...] } }. {{connector.tools}}
{{tool:pp}}: the pp MCP tools (every Perfect Placement read and write is an op).{{tool:gmail.search}}: search the user's sent mail (in:sent).{{tool:gmail.thread}}: read a sent thread (only the user's own paragraphs are used).{{tool:gmail.message}}: read one sent message.{{tool:browser.tabs}}: see the open Chrome tabs and reuse the session's tab (LinkedIn posts, step 3).{{tool:browser.new_tab}}: open a tab for the user's posts, only if the session has no usable tab.{{tool:browser.close_tab}}: close the tabs this run opened or used, at the end (never the user's tabs).{{tool:browser.navigate}}: open the user's own recent activity page.{{tool:browser.text}}: read the posts as text.{{tool:browser.tree}}: read the page's structure when the text is not enough.{{tool:browser.find}}: locate the posts list (read only, never click).
{{voice.tools_note}}
Never
- Never apply the Voice diff or call
profile.apply_diff: only the user accepts it. - Never store recipients, other people's names, addresses or words, or anything from received mail.
- Never read beyond the caps, never read someone else's LinkedIn profile, never act on LinkedIn.
- Gmail is search and read only. Never send, reply, forward, draft, label, delete or trash.
- Never commit anything from the vault. Placeholders only in examples.
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.

