linkedin

name	linkedin
description	Manage a LinkedIn company page — publish posts with consistent brand voice, grow engagement, respond to comments, analyze performance, preview profiles, and aggregate engagement for the monday dispatcher. Maintains a knowledge base of content ideas, audience insights, and brand guidelines. Use when the user wants to publish a LinkedIn post, draft LinkedIn content, check LinkedIn comments, respond to engagement, plan content strategy, review brand voice, view a LinkedIn profile, or manage a company page. Activate on mentions of LinkedIn, LinkedIn post, company page, social media, content calendar, engagement, brand voice, or audience growth.
allowed-tools	bash

LinkedIn Company Page Management

Complete LinkedIn presence management for any company page. Combines API automation with content strategy, brand voice, and engagement growth.

Configure your page with linkedin setup <companyId> --name="Your Company".

Quick start

# Post a text update
linkedin post "Excited to announce our latest open source contribution! #AIEcoverse"

# List recent posts with engagement stats
linkedin list --limit 5

# View and respond to comments
linkedin comments 7463311119181312000
linkedin comment 7463311119181312000 "Thanks for the feedback!"

# Profile lookup (before responding to someone)
linkedin profile klimetschek

# Watch for new comments (delegates to a scoop)
linkedin watch --scoop=linkedin-responder

# Monday aggregation
linkedin monday --limit 20 --date 3d

Brand Voice & Content Strategy

The skill maintains a brand knowledge base at /workspace/skills/linkedin/brand/:

Brand Voice (brand/voice.md)

Defines how your company speaks on LinkedIn. Updated by the user or inferred from successful posts. The voice guide covers:

• Tone and register (technical but approachable)
• Vocabulary preferences and avoidances
• Formatting conventions (emoji usage, hashtag strategy, line breaks)
• Content pillars (what we talk about and why)

When drafting posts, always read brand/voice.md first and conform to it.

Content Pillars (brand/pillars.md)

The recurring themes that define your LinkedIn presence:

• What topics we cover
• What angle we take on each
• What we never post about
• How we differentiate from generic AI hype

Audience Insights (brand/audience.md)

Who follows and engages with your page:

• Demographics and roles observed from commenters/reactors
• What content gets the most engagement
• Which profiles are VIPs (frequent engagers, industry voices)
• Engagement patterns by time/day

Content Log (brand/content-log.md)

Append-only record of every post published through the skill:

• Date, post text, engagement metrics (updated periodically)
• What worked, what didn't
• Ideas for future posts

Posting Cadence (brand/cadence.md)

Defines the rhythm:

• Target posting frequency (e.g., 3x/week)
• Best times to post (learned from engagement data)
• Content mix ratio (e.g., 40% technical, 30% community, 20% announcements, 10% culture)
• Scheduled drafts queue

Knowledge Base Operations

Building the brand KB

When the user first sets up the LinkedIn skill or asks to establish brand voice:

1. Analyze existing posts — linkedin list --limit 50 to understand current voice
2. Create brand/voice.md — synthesize the voice from existing content + user input
3. Create brand/pillars.md — define content themes
4. Create brand/audience.md — initialize from current followers/engagers
5. Create brand/cadence.md — set posting rhythm
6. Create brand/content-log.md — start tracking

Maintaining the KB

After every post:

• Append to brand/content-log.md with date and text
• After 48h, update with engagement metrics

Periodically (weekly or on request):

• Analyze what's working → update brand/audience.md
• Refine voice if the user gives feedback → update brand/voice.md
• Rotate content pillars if needed → update brand/pillars.md

Using the KB when drafting

When asked to draft or post:

1. Read brand/voice.md for tone
2. Read brand/pillars.md for topic alignment
3. Read brand/cadence.md to check if it's the right time
4. Read recent entries in brand/content-log.md to avoid repetition
5. Draft in the established voice
6. Present for approval before posting

Authentication

Current: Session-based (Voyager API)

Uses the active LinkedIn browser session (cookie + CSRF token). The LinkedIn tab must be open. Works immediately, no setup required.

Future: OAuth (Official API)

For headless operation (no browser tab needed), register a LinkedIn Developer App and request the Community Management API product. Then:

linkedin auth setup --client-id=<ID> --client-secret=<SECRET>

This enables:

• Posting without a LinkedIn tab open
• Stable, versioned API endpoints
• Proper rate limits and error handling
• Image/video upload support
• Scheduled posts via cron without browser dependency

Available commands

linkedin post <text> [--video=<path> | --image=<path> [--alt="..."]]

Publish a post to the configured company page.

• Posts as the company page (not personal profile)
• Visibility: Anyone (public)
• Include #hashtags and https://links inline
• With --video=<path>, attaches a video (mp4, H.264). Equivalent to linkedin video <path> "<text>". Uses the Voyager media-upload pipeline (register → PUT bytes → createShare with media). SINGLE-part uploads only for now; LinkedIn switches to MULTIPART for very large videos (not yet supported — the helper raises a clear error in that case).
• With --image=<path>, attaches an image (jpg/jpeg/png/gif/webp). Equivalent to linkedin image <path> "<text>". Uses the same Voyager pipeline as video with mediaUploadType: IMAGE_SHARING. Add --alt="<altText>" to set the image's accessibility alt text. Cannot be combined with --video.

linkedin video <path> "<text>"

Alt form of linkedin post --video=. Posts a video update with the given caption.

linkedin image <path> "<text>" [--alt="..."]

Alt form of linkedin post --image=. Posts an image update with the given caption and optional alt text.

linkedin list [--limit N]

List recent posts with engagement statistics (comments, likes, reposts).

linkedin comments <activityId>

View comments on a specific post.

linkedin comment <activityId> <text>

Add a top-level comment on a post as the company page. (Does not reply to a specific comment — it posts a new comment on the activity thread.)

linkedin reactions <activityId>

View who reacted to a post.

linkedin profile <vanityName|memberUrn>

Quick preview of a LinkedIn profile. Returns name, headline, location, summary, and current positions. Use before responding to commenters to understand context.

linkedin watch --scoop=<name> [--interval="cron"]

Start watching for new comments. Polls every 5 minutes (configurable), fires a webhook to the specified scoop when new comments arrive.

linkedin unwatch

Stop the comment watcher.

linkedin watches

Show current watch configuration and last-check status.

linkedin monday [--limit N] [--date Nd]

Monday protocol aggregation. Produces JSON array of:

• Posts with new engagement
• Individual new comments needing response

Watching for new comments

linkedin watch --scoop=linkedin-responder

Webhook payload on new comment:

{
  "type": "linkedin-comment",
  "event": "new-comment",
  "data": {
    "activityUrn": "urn:li:activity:...",
    "postText": "Original post (truncated)...",
    "postUrl": "https://www.linkedin.com/feed/update/...",
    "commentText": "The comment text",
    "commenter": "Commenter Name",
    "commentUrn": "urn:li:fsd_comment:..."
  }
}

The receiving scoop should:

1. linkedin profile <commenter> — understand who's commenting
2. Read brand/voice.md — respond in brand voice
3. linkedin comment <activityId> <text> — reply

Engagement Growth Playbook

Built into the brand KB, but the key principles:

1. Respond to every comment within 4 hours — signals the algorithm to boost
2. Ask questions in posts — drives comment engagement
3. Tag relevant people sparingly — only when genuinely relevant
4. Cross-reference with real conversations — share insights from actual work
5. Consistent cadence > viral attempts — steady posting beats sporadic bursts
6. Engage on others' posts — builds reciprocal relationships
7. Use the hook-body-CTA pattern — first line hooks, body delivers, end asks

How it works

The script:

1. Finds an open LinkedIn tab (or opens one)
2. Extracts the CSRF token from the JSESSIONID cookie
3. Makes API calls via playwright-cli eval from the page context
4. Requests carry cookies and correct Origin automatically

Endpoints reference

See references/endpoints.md for the full API documentation.

Limitations

• Both image and video posts work via the session-based Voyager pipeline (linkedin post --image=... / --video=...); OAuth is not required
• Large media files that LinkedIn marks as MULTIPART chunked uploads are not yet supported (SINGLE-part only — covers everything ~3-5 MB and short clips)
• Multi-image carousel posts (up to 20 images) are not yet supported — only one media item per post
• Click-to-tag-people on images (tapTargets) is not yet supported
• The Voyager API is undocumented and may change; queryIds are version-pinned
• Rate limits are unknown; use reasonable intervals for polling
• Profile lookup by vanity name requires page navigation (slower than URN lookup)
• Real-time notifications require polling (LinkedIn has no push for company pages)