Skip to content
Back to skills

Linkedin Reader

ASecurity

Read LinkedIn for financial research through opencli: the user's feed, and job search with location, seniority, job type, remote, company, and date filters. Use this skill whenever the user wants to check their LinkedIn feed, see what professionals or analysts are posting about a company, market, or earnings, gauge professional sentiment, or search finance and trading jobs on LinkedIn. Read-only: it cannot post, like, comment, connect, message, or apply.

  • 3,372 stars
  • 0 votes
  • 1 copy
  • 3 views
  • Added May 26, 2026
ai-agentsbashreactnodegitapidocumentation

Works with

  • terminal
  • cli
  • api

Security analysis

A96/100
  • mediumInstalls packages at runtime which could introduce malicious dependencies

Pro scans all 3 files and shows the line behind each finding

Scanned September 25, 2026

npx -y skills add himself65/finance-skills --skill linkedin-reader --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Linkedin Reader?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Linkedin Reader
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/himself65-linkedin-reader/badge)](https://www.skillsdirectory.com/skills/himself65-linkedin-reader)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: linkedin-reader
description: >
  Read LinkedIn for financial research through opencli: the user's feed, and job
  search with location, seniority, job type, remote, company, and date filters. Use
  this skill whenever the user wants to check their LinkedIn feed, see what
  professionals or analysts are posting about a company, market, or earnings, gauge
  professional sentiment, or search finance and trading jobs on LinkedIn. Read-only:
  it cannot post, like, comment, connect, message, or apply.
---

# LinkedIn Skill (Read-Only)

Reads LinkedIn for financial research using [opencli](https://github.com/jackwener/opencli), a universal CLI tool that bridges web services to the terminal via browser session reuse.

**This skill is read-only.** It is designed for financial research: reading professional commentary on markets, monitoring analyst posts, searching finance/trading jobs, and tracking professional sentiment. It does NOT support posting, liking, commenting, connecting, messaging, or any write operations.

**Important**: opencli reuses your existing Chrome login session — no API keys or cookie extraction needed. Just be logged into linkedin.com in Chrome and have the Browser Bridge extension installed.

---

## Step 1: Ensure opencli Is Installed and Ready

**Current environment status:**

```
!`(command -v opencli && opencli doctor 2>&1 | head -5 && echo "READY" || echo "SETUP_NEEDED") 2>/dev/null || echo "NOT_INSTALLED"`
```

If the status above shows `READY`, skip to Step 2. If `NOT_INSTALLED`, install first:

```bash
# Install opencli globally
npm install -g @jackwener/opencli
```

If `SETUP_NEEDED`, guide the user through setup:

### Setup

opencli requires Node.js >= 20 and a Chrome browser with the Browser Bridge extension:

1. **Install the Browser Bridge extension:**
   - Download the latest `opencli-extension-v{version}.zip` from the [GitHub Releases page](https://github.com/jackwener/opencli/releases)
   - Unzip it, open `chrome://extensions` in Chrome, and enable **Developer mode**
   - Click **Load unpacked** and select the unzipped folder
2. **Login to linkedin.com** in Chrome — opencli reuses your existing browser session
3. **Verify connectivity:**

```bash
opencli doctor
```

This auto-starts the daemon, verifies the extension is connected, and checks session health.

### Common setup issues

| Symptom | Fix |
|---------|-----|
| `Extension not connected` | Install Browser Bridge extension in Chrome and ensure it's enabled |
| `Daemon not running` | Run `opencli doctor` — it auto-starts the daemon |
| `No session for linkedin.com` | Login to linkedin.com in Chrome, then retry |
| `AuthRequiredError` | LinkedIn session expired — refresh linkedin.com in Chrome and log in again |

---

## Step 2: Identify What the User Needs

Match the user's request to one of the read commands below, then use the corresponding command from `references/commands.md`.

| User Request | Command | Key Flags |
|---|---|---|
| Setup check | `opencli doctor` | — |
| Home feed / posts | `opencli linkedin timeline` | `--limit N` (default 20, max 100) |
| Search for jobs | `opencli linkedin search "QUERY"` | `--location`, `--limit N` (default 10, max 100), `--details` |
| Finance job search | `opencli linkedin search "QUERY"` | `--experience-level`, `--job-type`, `--remote`, `--company`, `--date-posted`, `--start` |

---

## Step 3: Execute the Command

### General pattern

```bash
# Read LinkedIn feed posts
opencli linkedin timeline --limit 20 -f json

# Search for finance/trading jobs
opencli linkedin search "quantitative analyst" --limit 10 -f json
opencli linkedin search "portfolio manager" --location "New York" --limit 15 -f json

# Detailed job listings with descriptions
opencli linkedin search "financial analyst" --details --limit 10 -f json
```

### Key rules

1. **Check setup first** — run `opencli doctor` before any other command if unsure about connectivity
2. **Use `-f json` or `-f yaml`** for structured output when processing data programmatically
3. **Use `-f csv`** when the user wants spreadsheet-compatible output
4. **Use `--limit N`** to control result count — start with 10-20 unless the user asks for more
5. **For job search, use filters** — `--location`, `--experience-level`, `--job-type`, `--remote`, `--date-posted` to narrow results
6. **Read-only** — don't post, like, comment, connect, message, or apply to jobs

### Output format flag (`-f`)

| Format | Flag | Best for |
|---|---|---|
| Table | `-f table` (default) | Human-readable terminal output |
| JSON | `-f json` | Programmatic processing, LLM context |
| YAML | `-f yaml` | Structured output, readable |
| Markdown | `-f md` | Documentation, reports |
| CSV | `-f csv` | Spreadsheet export |

### Output columns

**Timeline** posts include: `rank`, `author`, `author_url`, `headline`, `text`, `posted_at`, `reactions`, `comments`, `url`.

**Job search** results include: `rank`, `title`, `company`, `location`, `listed`, `salary`, `url`. With `--details`: also `description`, `apply_url`.

---

## Step 4: Present the Results

Lead with what matters for the user's research question rather than a raw dump. For feed posts, attribute each one (author, headline, text, reactions and comments, URL), highlight market commentary, analyst takes, and earnings reactions, and note the bullish or bearish lean and where views are consensus or contrarian. For job results, give title, company, location, salary when available, posting date, and the listing URL. Keep browser session details private.

---

## Step 5: Diagnostics

If something isn't working, run:

```bash
opencli doctor
```

This checks daemon status, extension connectivity, and browser session health.

---

## Error Reference

| Error | Cause | Fix |
|-------|-------|-----|
| `Extension not connected` | Browser Bridge not installed/enabled | Install extension and enable it in Chrome |
| `No session` | Not logged into linkedin.com | Login to linkedin.com in Chrome |
| `AuthRequiredError` | LinkedIn login wall detected | Refresh linkedin.com and log in again |
| `EmptyResultError` | No results found for query | Broaden search terms or check feed has content |
| Rate limited | Too many requests | Wait a few minutes, then retry |

---

## Reference Files

- `references/commands.md` — Complete read command reference with all flags, research workflows, and usage examples

Read the reference file when you need exact command syntax, research workflow patterns, or output details.

Files in this skill

  • README.md2.1 KB
  • SKILL.md7.1 KB
  • references/commands.md7.6 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.