li-reply

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 itself connects to nothing — no LinkedIn API, no browser, no login, no cookie or session access. It works only on comments I paste into our chat myself (ideally with names and roles; screenshots are fine). I need my own LinkedIn account only in the ordinary sense that the comments come from my own LinkedIn post and I paste the finished replies back into LinkedIn myself; no credential is given to the skill. The pack's README describes an optional voice profile at ~/.claude/linkedin/voice.md (from templates/voice.md in the repository) that it says every skill in the pack reads, but li-reply's own SKILL.md never references that file, it is not part of this install, and the skill works from the pasted comments alone. Install "li-reply" for me. It handles the reply thread under my own LinkedIn posts: it sorts every comment I paste into five buckets — LEAD (someone describing the problem I solve), SUBSTANCE (adds data, disagrees, extends), PEER (a name worth being seen next to), SUPPORT ("great post", an emoji, a tag) and NOISE (pitch, spam, bad faith) — says the count in each bucket out loud, then drafts replies in that order (LEAD first, NOISE last or skipped), stopping when the value stops. Its reply rules: answer the actual question in the reply itself instead of sending people to a DM, use the commenter's name once at the start and never with an exclamation mark, match the comment's length, concede the true part first to a critic and never reply twice on the same thread, answer a lead fully in public with one offer-of-help sentence at the end at most, give SUPPORT a like and a 3-8 word reply at most, and ignore pitches in the comments. The output is one copy-ready block grouped by bucket, and nothing is posted until I say yes: I paste the replies into LinkedIn myself. Repository: https://github.com/avi691/linkedin-agent-skill (the skill is the whole skills/li-reply/ 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-reply/ 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 26f95e55f384ca2ec1b692d3f95d08d33796fea404db8740387aca15b1248751 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 (3,016 bytes). There is no script, no code, no data file and no other skill folder in the install perimeter. SKILL.md itself references no other file in the repository: it does not call the sibling /li-human skill, its humanize.py / detect.py scripts or its slop.json, and it does not reference templates/voice.md or ~/.claude/linkedin/voice.md. Its output section describes the replies as "already humanized", but the file gives no instruction to load or run any humanizer: treat that as a writing-style expectation inside this one file, not a file dependency. The pack-level facts, outside this skill's own text, are: the README says every skill in the pack reads ~/.claude/linkedin/voice.md if the user has set one up, and describes /li-human as a separate sibling skill with two local Python scripts; neither is installed by this install and li-reply does not require either to run as written. - 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 (there is no script in the folder), no key or account is asked for, and there is no telemetry in the file. SKILL.md contains no URL at all. - What happens at use time, disclosed here before it ever happens — using the skill is a separate step that only starts when I paste comments from my own LinkedIn post and ask for replies: the comments I paste (including the commenters' names and roles, if I include them, or a screenshot, if I paste one) are processed in our conversation only, to sort them into the five buckets and draft the replies. Nothing is sent to LinkedIn or to any other service by the skill: SKILL.md explicitly says "Do not scrape the thread with a browser tool", there is no LinkedIn API, no browser automation, no login, and no access to LinkedIn cookies, session data or any credential anywhere in the file. Nothing is posted automatically and no reply is sent in my name: SKILL.md's gate is "nothing is posted until the user says yes", and I paste the approved replies into LinkedIn myself, under my own account, outside the skill. Tell me this plainly in your summary, because a skill that answered comments in my name automatically would be a different product. - The local-only parts at use time: the triage (LEAD / SUBSTANCE / PEER / SUPPORT / NOISE counts, said out loud), the reply drafting under the rules above (answer in public, name once with no exclamation mark, matched length, critic and lead handling, pitches ignored), and the grouped copy-ready output block all stay between you and me in the chat. The skill invents no access: if I have not pasted a comment, it is not in scope, and screenshots I paste are the only other input SKILL.md accepts. - Do not scrape LinkedIn, do not open a browser tool on the thread, do not ask for my LinkedIn login, cookies or session, do not post or send any reply, and do not install any sibling skill (including /li-human) or copy templates/voice.md during this install: using the skill, and any pack-level setup like a voice profile, are my decisions, later, when I actually ask for replies. 3. When I say "yes, install", install it as a skill: copy the whole li-reply folder into the agent's skills directory, in a folder named "li-reply", keeping its 1 file (SKILL.md at its root). A manual clone followed by copying the skills/li-reply 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-reply), the 1 file in place with matching fingerprint. Do not run anything to verify (there is nothing runnable in the folder). 5. Report what was installed, where, and what I still need to do myself (for example: paste the comments from my own LinkedIn post, ideally with names and roles, or a screenshot, when I want replies; review the draft replies bucket by bucket; paste the ones I approve into LinkedIn myself — nothing is ever posted for me; optionally, as a separate pack-level step that this install does not do, set up a voice profile from templates/voice.md in the repository if I want one; using the skill is a separate step that only starts when I ask for it, and no sends happen at use time either, beyond my own manual posting on LinkedIn). 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.