linkedin-export

这段内容是审核时保存的快照。外部来源才是完整且最新的版本。

---
name: linkedin-export
description: "Parse, search, analyze, and ingest LinkedIn GDPR data exports into structured JSON or RLAMA for semantic search. Covers messages, connections, profile data, and Markdown export. Requires a LinkedIn GDPR ZIP file. Triggers on 'LinkedIn data', 'search messages', 'analyze connections', 'LinkedIn export', 'GDPR download'."
---

# LinkedIn Export Skill

Parse LinkedIn GDPR data exports into structured JSON, then search messages, analyze connections, export to Markdown, and ingest into RLAMA for semantic search.

## Prerequisites

- **Python 3.10+** via `uv`
- **LinkedIn GDPR export ZIP** — Request at: LinkedIn → Settings → Data Privacy → Get a copy of your data
- **RLAMA + Ollama** (optional, for semantic search ingestion)

## Quick Start

```bash
# 1. Parse the export ZIP (run once)
uv run ~/.claude/skills/linkedin-export/scripts/li_parse.py ~/Downloads/Basic_LinkedInDataExport_*.zip

# 2. Search, analyze, export, or ingest
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --list-partners
uv run ~/.claude/skills/linkedin-export/scripts/li_network.py summary
uv run ~/.claude/skills/linkedin-export/scripts/li_export.py all --output ~/linkedin-archive/
uv run ~/.claude/skills/linkedin-export/scripts/li_ingest.py
```

All scripts read from `~/.claude/skills/linkedin-export/data/parsed.json`. Parse once, query many times.

---

## Parse — `li_parse.py`

Unzip and parse all CSVs from the LinkedIn GDPR export into structured JSON.

```bash
uv run ~/.claude/skills/linkedin-export/scripts/li_parse.py <linkedin-export.zip>
uv run ~/.claude/skills/linkedin-export/scripts/li_parse.py <zip> --output /custom/path.json
```

**Output**: `~/.claude/skills/linkedin-export/data/parsed.json`

Parses 23 CSV types:

**Core**: messages, connections, profile, positions, education, skills, endorsements, invitations, recommendations, shares, reactions, certifications

**Extended**: comments (548), projects (3), honors (2), organizations (3), volunteering (1), languages (9), events (12), member_follows (828), job_applications (443, merged from multiple files), recommendations_given (3), inferences (4)

Auto-detects CSV column names (case-insensitive), handles LinkedIn's preamble format (Connections.csv), and merges split files (Job Applications).

---

## Search Messages — `li_search.py`

Search messages by person, keyword, date range, or combination.

```bash
# Search by person
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --person "Jane Doe"

# Search by keyword
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --keyword "project proposal"

# Date range
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --after 2025-01-01 --before 2025-06-01

# Combined filters
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --person "Jane" --keyword "meeting" --after 2025-06-01

# Full conversation by ID
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --conversation "CONVERSATION_ID"

# List all conversation partners (sorted by message count)
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --list-partners

# Show context around matches
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --keyword "AI" --context 3

# Full message content + JSON output
uv run ~/.claude/skills/linkedin-export/scripts/li_search.py --keyword "proposal" --full --json
```

**Flags**: `--person`, `--keyword`, `--after`, `--before`, `--conversation`, `--list-partners`, `--context N`, `--full`, `--limit N`, `--json`

---

## Network Analysis — `li_network.py`

Analyze the connection graph — companies, roles, timeline.

```bash
# Summary stats
uv run ~/.claude/skills/linkedin-export/scripts/li_network.py summary

# Top companies by connection count
uv run ~/.claude/skills/linkedin-export/scripts/li_network.py companies --top 20

# Connection timeline
uv run ~/.claude/skills/linkedin-export/scripts/li_network.py timeline --by year
uv run ~/.claude/skills/linkedin-exp