Li Comment

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, and nothing else to install first. For using the skill afterwards, read this first: the skill only writes comments, it never posts them — posting a comment on LinkedIn is something I do myself, by hand, under my own LinkedIn account, and the skill never touches that account, its credentials, its cookies or its session (see point 2). There is no LinkedIn API, no key and no paid service at use time either: I paste the post text myself, and the drafts come back to me in this chat. One cross-skill dependency is written into SKILL.md: before showing me the two comment options, it says to run both through /li-human, the humanizer skill from the same pack (skills/li-human/, with its own SKILL.md plus the Python tools humanize.py and detect.py and the slop.json lexicon). /li-human is a separate skill and is NOT installed by this install: without it, that humanizing pass cannot run as written. The pack also has an optional voice profile (templates/voice.md in the repository, copied by the user to ~/.claude/linkedin/voice.md) which the repository README says every skill in the pack reads; li-comment's own SKILL.md does not name that file, and this install does not create it. Install "li-comment" for me. It writes comments on other people's LinkedIn posts that read as a person with an opinion, not a bot. From a post I paste (text, plus the author's name and role if I have them, or a screenshot it reads), it picks one of nine comment types by what the post actually is — add a datum, add the missing case, respectful disagree, extend one line, ask the real question, the receipt, the correction, the reframe, or the one-liner (under 12 words) — never defaulting to type 1, and drafts under strict rules: 2 to 4 sentences, one idea, the specific thing, never opening with "Great post", "Love this", "So true", "Couldn't agree more", "This resonates" or the author's first name with an exclamation mark, no emoji opener, never restating the post, and disagreement allowed only after a real agreement. It returns two options of different types, labelled, plus a one-line reason for the one it would post, after running both through /li-human first. In batch mode, for an engagement round, it takes 5-10 posts pasted in one message and returns one comment each in a single block. Repository: https://github.com/avi691/linkedin-agent-skill (the skill is the whole skills/li-comment/ folder at the pinned commit: SKILL.md only, 1 file) Pinned version: install the code at commit e4ec4fb6f1b0e227eb9b14b31b735072db5df9cb exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/avi691/linkedin-agent-skill into a temporary folder, check out commit e4ec4fb6f1b0e227eb9b14b31b735072db5df9cb, and copy the skills/li-comment/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 05bfb1fb28976d6ee5d50b31ec7c8d7fc13fba9db71257dc1dad2a0ffd21e071 If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, and tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the file: - The skill folder holds 1 file only: SKILL.md (4,160 bytes). There are no scripts, no data files and no code to run in this folder. The one dependency outside the folder is named in SKILL.md itself: "Run both through /li-human first" — /li-human lives in skills/li-human/ in the same repository (SKILL.md plus humanize.py, detect.py and slop.json), it is not part of this folder and this install does not copy it. Batch mode also names one local file outside the folder: a running note of who I have already commented on this week, kept in ~/.claude/linkedin/log.md on this machine. SKILL.md names no other file from the repository — not the other 10 skills, not the README; the repository README's statement that every pack skill reads ~/.claude/linkedin/voice.md is pack-level context, and li-comment's own SKILL.md does not instruct reading it. - Nothing is sent anywhere during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. No script is run during installation (the folder contains none), no key is asked for, and there is no telemetry and not a single URL in the folder. - What leaves this machine at use time, disclosed here before it ever happens — using the skill is a separate step that only starts when I paste a post and ask for a comment: the skill itself sends nothing anywhere. It connects to no service, calls no API and opens no website. The post text I paste (and the author's name and role if I give them, or a screenshot I provide, which it reads) stays between you and me in this chat, under this agent's own terms, and the two comment options come back the same way. It never reads or asks for my LinkedIn credentials, cookies or session, and nothing in the folder or the repository accesses them. It never posts: SKILL.md says "Do not auto-post. Do not use a browser tool to publish comments on the user's behalf", it forbids using browser automation to scrape the feed, and if I give it a URL it cannot open it must ask me to paste the text instead of guessing what the post said. The only send to LinkedIn is my own, later and outside the skill: I copy the comment I choose and paste it on LinkedIn myself, by hand, under my own account. In batch mode the only write at use time is local: the running note appended to ~/.claude/linkedin/log.md on this machine — nothing is uploaded. If I separately install /li-human and it runs on a draft, its tools run locally on my text and upload nothing (that is what its SKILL.md and the repository README state); that is /li-human's behaviour, not part of this install, and it never happens during this installation. - The local-only parts at use time: everything the skill does is local drafting between you and me — picking the comment type, applying the rules (2 to 4 sentences, one idea, no banned openers, no emoji opener, no restating the post), writing the two labelled options and the one-line recommendation. No number, client or outcome is invented for me: types like "add a datum" and "the receipt" only work with facts I actually gave. - Do not install /li-human or anything else during this install, do not create ~/.claude/linkedin/voice.md or ~/.claude/linkedin/log.md, do not open LinkedIn, do not use any browser tool, and do not write or post any comment: using the skill is my decision, later, when I actually paste a post. 3. When I say "yes, install", install it as a skill: copy the whole li-comment folder into the agent's skills directory, in a folder named "li-comment", keeping its 1 file (SKILL.md at its root). A manual clone followed by copying the skills/li-comment folder into ~/.claude/skills/ is the route described here; any other route is fine only as long as every installed file matches its fingerprint. 4. Verify with no network calls: frontmatter valid (name: li-comment), the 1 file in place with matching fingerprint. There are no scripts to run for verification, and do not run any. 5. Report what was installed, where, and what I still need to do myself (for example: paste the post text myself, with the author's name and role if I have them, whenever I want a comment, and expect two labelled options of different types plus a recommendation; post the comment I choose myself, by hand, on LinkedIn under my own account, since the skill never posts, never scrapes and never touches my account; optionally, install /li-human separately if I want the humanizing pass SKILL.md calls for, and optionally fill in the pack's voice profile at ~/.claude/linkedin/voice.md from templates/voice.md — neither is done by this install; using the skill is a separate step that only starts when I ask for it, and the use-time behaviour disclosed above — no sends by the skill, a local log note in batch mode only — happens only then). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.