import-linkedin-history

name	import-linkedin-history
description	Parse a LinkedIn data export ZIP — positions, connections, messages, profile — and land structured markdown/CSV in <WORKING_FOLDER>/linkedin/. Surfaces follow-up candidates (recent connections + recent message threads) for the user to consider routing to outreach.
disable-model-invocation	false
allowed-tools	Read, Write, Bash(unzip *), Bash(mkdir *), Bash(test *), Bash(ls *), Bash(file *), Bash(head *), Bash(wc *)

Import LinkedIn History

Accept a LinkedIn data export ZIP, parse it, land structured outputs. Optionally cross-reference connections against the user's target lists.

Inputs

$ARGUMENTS:

• Required: <zip-path> — path to the LinkedIn export ZIP (the "Get a copy of your data" download).
• Optional: --cross-ref=<domain-slug> — flag connections at companies in this domain note.
• Optional: --surface-followups — surface recent connections / message threads as outreach candidates.

Examples

/career:ecosystem linkedin-import ~/Downloads/Basic_LinkedInDataExport_2026-04-28.zip
/career:ecosystem linkedin-import ~/Downloads/li-export.zip --cross-ref=ai-eval-startups --surface-followups

Procedure

1. Validate + extract

Check ZIP exists. Extract to ${WORKING_FOLDER}/linkedin/raw/ (overwrite on re-run with confirmation).

LinkedIn exports typically contain: Profile.csv, Positions.csv, Education.csv, Connections.csv, messages.csv, Endorsement_Received_Info.csv, Skills.csv, Recommendations_Received.csv (exact filenames vary by export tier).

Inventory the files present and the row counts.

2. Parse + structure

Land parsed outputs:

• ${WORKING_FOLDER}/linkedin/profile.md — single document with profile + positions + education + skills + endorsements summary.
• ${WORKING_FOLDER}/linkedin/connections.csv — normalized columns: name, headline, company, role, connected_on, profile_url.
• ${WORKING_FOLDER}/linkedin/messages/<thread-id>.md — one file per thread (or one consolidated messages.md for sparse exports).

3. Cross-reference (if --cross-ref)

For each connection, check whether their company matches a company in domain-notes/${cross-ref}.md. Tag matches in ${WORKING_FOLDER}/linkedin/cross-ref-${cross-ref}.md:

# LinkedIn × {{domain}} overlap

| connection | role | company (matched) | connected_on | last_message_at | suggested action |

4. Surface follow-ups (if --surface-followups)

• Recent connections (last 90 days) with no follow-up message → list as ## Cold connections — never followed up.
• Active message threads with the user as last sender > 14 days ago → list as ## Stalled threads — your move.
• Active threads with the counterparty as last sender → list as ## Threads waiting on you.

Write to ${WORKING_FOLDER}/linkedin/follow-ups.md. Don't auto-route to outreach.

5. Print summary

linkedin imported: linkedin/
profile + positions + connections + messages: parsed
connections: <n>
messages: <n> threads
cross-ref overlaps: <n> (if --cross-ref)
follow-up candidates: <n> (if --surface-followups)

next:
  cross-ref → /career:outreach find <company> (already-connected? skip Hunter)
  follow-ups → review linkedin/follow-ups.md, route to /career:outreach draft for the strongest 2–3

Guardrails

• Don't write back to LinkedIn. This skill parses an export — it doesn't post, message, or connect.
• Connections data is private. Files land in ${WORKING_FOLDER}/linkedin/ which the workspace .gitignore should already cover via linkedin/ (add it if not).
• When suggesting follow-ups, don't draft messages — that belongs to draft-outreach. This skill surfaces candidates only.

Failure modes

• ZIP doesn't look like a LinkedIn export → bail with a list of files found vs. expected filenames.
• Some columns missing (LinkedIn export tier varies) → continue with the columns present; mark missing fields explicitly.
• No connections file in export → write profile-only output and warn.