linkedin-comment
Skill0 starsChanged 43 days agoMITView on GitHub
Comment on someone else's LinkedIn post, reshare a post with commentary, or find posts worth commenting on. Use when the user pastes a post or its URL and wants a comment or a repost with their take, or asks where to comment today. Drafts 1 to 3 variants plus a reaction in the persona's voice and checks the log so no post gets a second comment. Not for answering comments on the user's own post or replies to their comments (use linkedin-reply).
2 other projects use a version of this.
Adopted by
Most recently taken up by
- matteotitta/claude-marketing-linkedintheir own version, first seen Oct 25, 2026
- cavalenzuelalara/claude-skillstheir own version, first seen Oct 11, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- cavalenzuelalara/claude-skills★ 0same name
- matteotitta/claude-marketing-linkedin★ 0same name
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.
In this repository9 other files it gives its agents.
- linkedin-dmSkill
- linkedin-humanizeSkill
- linkedin-inboxSkill
- linkedin-planSkill
- linkedin-postSkill
- linkedin-profileSkill
- linkedin-replySkill
- linkedin-setupSkill
- linkedinSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Log: Read the persona's log per ../linkedin/references/log-format.md ("Reading").
- Very specific: line 25
Context: Read the whole post and its top comments (../linkedin/references/guardrails.md rule 1).
- Very specific: line 28
Stop checks: Stop and tell the user when: the log shows a comment on this post already (offer a reply in that thread via linkedin-reply …
- Very specific: line 60
Guardrails: Run ../linkedin/references/guardrails.md over every variant (it runs ../linkedin/references/voice.md).
- Very specific: line 69
Log: After the send or the user's "posted", append a comment line (to: post URL, name: the post's author, text: the exact comment) and a …
- Distinctive: line 76
Read the post (draft step 1) and check the log for an earlier reshare of it.
- Distinctive: line 77
Ask whether the user wants commentary.
- Very specific: line 107
Follow ../linkedin/SKILL.md "After a good run".
≈ “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
Comment
A comment is a turn in someone else's conversation. It earns its place by adding one thing the post and the comments above it do not have (C1). Three modes: draft, reshare, find. Find ends in draft.
Before any mode
- Persona. Load it per
../linkedin/references/persona-template.md("Where it lives"). Done when you know the persona (or that none exists and the neutral defaults apply) and have read its Voice, Story bank, Off limits and Offer sections. - Log. Read the persona's log per
../linkedin/references/log-format.md("Reading"); a log with no comment history gets seeded first ("Seeding"). Done when you know, for every candidate post: whether this persona or another of the user's personas already commented on it, and which authors this persona commented on today.
Draft
For a post the user pasted, linked or picked in find.
- Context. Read the whole post and its top comments (
../linkedin/references/guardrails.mdrule 1); with only a URL and no Curviate, ask for a paste and offer the fetch once (../linkedin/references/curviate.md). Done when you have the post text and the top comments, or the user said there are none. - Stop checks. Stop and tell the user when: the log shows a comment on this post already (offer a reply
in that thread via
linkedin-replyinstead), or another persona of the user is already in this thread (two of one person's accounts in one thread reads as a pod, R6, R13). A second angle on a post you already commented on becomes a reaction, not a second comment. When the persona already commented on this author today, say so and suggest a reaction or waiting a day (one author a day, seereferences/choosing-posts.md). Pick the angle. Name the one thing the post and its comments are missing, then choose the angle type that delivers it. Done when you can say the addition in one line and it traces to the persona's story bank, the post itself, or something the user said this session (guardrails rule 2). When the persona has nothing real to add, say so and suggest a reaction only.
Angle Use when Shape Missing piece the post is right but leaves out a condition, cost or step the persona has seen name the piece, then the case where it decides the outcome Answer their question the post ends on a real question answer first in one sentence, then one concrete example Example from experience the persona has a number, case or result on this exact topic the specific thing, what it measured, what it changed (W4) Counter with concession the persona disagrees with part of it grant the point that holds in their terms, then the narrow disagreement and its reason, hedged as the persona's view (C3) Sharper question the post leaves its hardest part open one question about what the author actually said, that the persona wants answered (C2) - Write 1 to 3 variants, each a different angle when more than one fits, each 2 to 4 sentences: count
them before the card. Shape per
../linkedin/references/voice.md("Per-surface shape", Comment): a reply in a conversation, first person, opening on their claim with the persona's take. Close on a take or a real question. Address the author by first name only when the persona does. A run of five or more words from the persona's recent comments in the log gets rewritten. Within one batch, each story-bank fact or anecdote is used at most once, and variants of one post use different material. - Own offer. By default the comment leaves out the persona's offer and affiliations (persona "Offer"). A mention fits when the persona's "when it is fair to mention" allows it, the thread is about the problem or the category of tools the offer belongs to, and the offer genuinely answers what is being discussed. Then one plain sentence, no link, disclosed as the persona's own or as an affiliation ("I built X", "I work with X"). Undisclosed praise of an affiliated offer is what R16 and the persona's credibility both rule out.
- Reaction. Pick one reaction that fits the post: Insightful for an idea the persona learned from, Support for a hard moment, Celebrate for news or a win, Love for something personal and warm, Funny only for intended humour, Like otherwise. A reaction alone is the right call when there is nothing to add.
- Guardrails. Run
../linkedin/references/guardrails.mdover every variant (it runs../linkedin/references/voice.md). Done when every variant passes every rule and the language matches the post (rule 4). - Approval card. One card per rule 7, the variants numbered, each with its angle and a "why" line, the reaction on its own line. The user picks a variant or edits.
- Send or hand off. With Curviate connected, react, then comment, one write at a time (guardrails rule 8,
../linkedin/references/curviate.md"Sending through it"). Otherwise the card is the hand-off. Onautopilot(preconditions in../linkedin/references/curviate.md"Autopilot"), say once per session before the first unattended comment that LinkedIn prohibits automated commenting and hides such comments (R10), the riskiest use of autopilot; then continue as the persona chose. - Log. After the send or the user's "posted", append a
commentline (to: post URL,name: the post's author,text: the exact comment) and areactionline. Done when both lines are written.
Reshare
For "repost this", "share this with my network", "reshare with my take".
- Read the post (draft step 1) and check the log for an earlier reshare of it.
- Ask whether the user wants commentary. A plain repost is a valid choice; skip to step 4.
- Draft the commentary: one to three sentences on why this is worth the reader's time, carrying the persona's own angle or experience (H10), never a summary of the post. The author is tagged only when the persona wants them notified (R22). Run guardrails.
- Approval card with the post URL and the commentary (or "plain repost").
- Hand-off: the user uses LinkedIn's "Repost" (with or without thoughts). When LinkedIn shows no repost option,
the author turned resharing off; tell the user. Send through Curviate only if its CLI lists a repost command
(
curviate post --help). - Log a
postline withnote: "reshare of <post URL>"and the commentary astext.
Find
For "where should I comment today", "find posts on X", or a list of candidate posts.
- Gather candidates. Outside the persona's
hours, say first that approved comments wait for the next window while the posts age, and offer to search then instead. With Curviate connected (reads per../linkedin/references/curviate.md):search posts --verboseon phrases the persona's audience actually uses (practitioner wording, not bare platform terms) andfeed home, dropping items without text. Otherwise ask the user to paste candidates (URL plus text), and offer Curviate once. Done when you have at least a handful of candidates or the user's full list. - Qualify. Apply
references/choosing-posts.md(read budget included) to each candidate, reading about 10 comments on each one that survives the first cuts. Done when every candidate is marked keep or skip with a reason. - Shortlist. Show a table: author, post (first line), age, comment count, audience fit, suggested angle, and the skip reasons as counts ("9 found: 3 kept, 2 lead magnets, 1 job post, 2 already commented, 1 same author as yesterday"). The user picks; when they already asked for drafts, take the top 2 or 3 keepers.
- Draft each picked post with the Draft mode, one comment per post, the cards below the table.
After a good run
Follow ../linkedin/SKILL.md "After a good run".
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.

