Linkedin Outreach

LinkedIn Outreach

Build a LinkedIn outreach campaign. The agent writes the personalized sequences (via linkedin-message-writer); the bundled scripts lint char limits, export the tool CSV, and optionally launch a Phantombuster send. Default path is the CSV export (LinkedIn tools are browser-based importers).

When to use

• "LinkedIn outreach" / "set up a LinkedIn campaign" / "connect with these leads on LinkedIn."
• The LinkedIn counterpart to cold-email-outreach; a launch target for the signal composites.

How it works (agent + scripts)

1. (agent) Intake goal, angle, name, tool, lead selection, sequence config, account tier.
2. (agent) Load leads (store query / CSV / upstream output).
3. (agent) Write per-lead sequences (CR + follow-ups + optional InMail), enforcing limits (CR 200/300, message 8000, InMail subj 200 / body 1900). Lint them; rewrite over-limit.
4. (agent) Present samples for review; iterate.
5. Export the tool CSV — or launch via Phantombuster (automated send).
6. (agent) Log the campaign + contacts; dedup against the outreach log.

How to run

Lint the drafted sequences (message carries the CR / DM text; per-type limits enforced):

python3 ${SKILL_DIR}/scripts/linkedin_lint.py --input ${WORKSPACE}/sequences.json

Export a tool-ready CSV (default path — nothing is sent):

python3 ${SKILL_DIR}/scripts/export_csv.py --input ${WORKSPACE}/sequences.json \
  --tool dripify --output ${WORKSPACE}/linkedin_campaign.csv

Optional automated send via Phantombuster (needs key + LinkedIn session cookie; throttle):

python3 ${SKILL_DIR}/scripts/phantombuster_send.py \
  --agent-id 1234567890 --leads ${WORKSPACE}/sequences.json \
  --daily-cap 50 --dry-run            # drop --dry-run to actually launch

export_csv --tool: generic | dripify | expandi | botdog | phantombuster.

Outputs

• A tool-ready CSV (linkedin_url, first_name, last_name, company, title, connection_request, followup_1..N, inmail_subject, inmail_body) — or a Phantombuster launch result — plus an outreach-log record for the store.

Credentials / env

• env.required: none. The default CSV-export path needs no key (the agent writes the copy; lint + export are stdlib).
• env.optional (all degrade): Send — if PHANTOMBUSTER_API_KEY (+ LINKEDIN_SESSION_COOKIE) is set → automated LinkedIn send; else → export a CSV to import manually (the keyless default). APIFY_API_TOKEN — if set → cookieless Apify research via linkedin-profile-post-scraper; else → that skill's keyless Playwright li_at-cookie degrade. Store creds (SUPABASE_*) — if set → read/log; else → CSV/paste.

Notes & edge cases

• LinkedIn tools are browser-based CSV importers — the default path produces a CSV, not an API send. The Phantombuster path needs a key + session cookie and must be throttled (--daily-cap) to dodge LinkedIn limits/bans.
• Strictly enforce char limits; rewrite to fit, never truncate. Always show samples before exporting a large batch.
• Dedup against the outreach log so the same lead isn't re-sequenced.