Linkedin Comment To Outreach

Steps

<!-- houston-workflow:v1 -->

1. Validate inputs. Check the URL is a LinkedIn post (not a profile, not an article), confirm the four Composio connections, read config/context-ledger.json. Mint a runId of the form {YYYY-MM-DD}-{post-slug} and create runs/{runId}/notes.md for the per-run journal.

2. Phase 1 - Scrape commenters. Call linkedin-comment-scraper with the post URL. Result lands at runs/{runId}/scrape.json. Append summary to runs/{runId}/notes.md.

   Checkpoint. Tell the user: "Scraped {N} unique commenters from {author}'s post. Moving on to Airtable."

3. Phase 2 - Load to Airtable. Call airtable-lead-loader with runs/{runId}/scrape.json and the chosen base ID. The table name is LinkedIn Commenters - {author} - {YYYY-MM-DD}. Append summary to runs/{runId}/notes.md with the table ID and the load count.

   Checkpoint. Tell the user: "Loaded {N} records into Airtable. Starting Apollo enrichment."

4. Phase 3 - Enrich with Apollo. Call apollo-enrichment with the Airtable base + table ID. Result: Airtable rows updated with email + company + title + location, and Apollo contacts created under label LinkedIn Comments - {author} Post. Re-fetch the rows that came back with a verified email and save to runs/{runId}/contacts.json. Append match-rate summary to runs/{runId}/notes.md.

   Checkpoint. Tell the user: "Found emails for {M} of {N} commenters ({M/N}% match rate). {M} contacts ready for outreach. Moving on to the email sequence."

5. Phase 4 - Co-write the sequence. Call cold-email-sequence. This is the interactive phase - I work with you one email at a time, locking each before moving to the next. Save to sequences/{runId}-sequence.md. Append the lock summary to runs/{runId}/notes.md.

   Checkpoint. Tell the user: "Sequence locked. Loading into Instantly."

6. Phase 5 - Create the Instantly campaign. Call instantly-campaign with sequences/{runId}-sequence.md and runs/{runId}/contacts.json. The campaign name is LinkedIn - {author} - {short topic}. Always paused. Append the Instantly campaign ID and lead-load summary to runs/{runId}/notes.md. Append a row to campaigns.json with status: "paused".

7. Final summary. One short block to the user:

   • Campaign name + status (paused).
   • Lead count loaded.
   • Sending accounts attached.
   • Schedule (Mon-Fri, 8-5 in your default timezone).
   • "Review in Instantly. Activate when you're ready - I won't do it for you."

Outputs

• runs/{runId}/scrape.json - deduped commenter list from phase 1.
• runs/{runId}/contacts.json - Apollo-enriched contacts with verified emails (drives the Instantly load).
• runs/{runId}/notes.md - per-run journal with checkpoints, counts, and decisions.
• sequences/{runId}-sequence.md - locked 3-email sequence.
• New Airtable table LinkedIn Commenters - {author} - {date} populated with the full lead-tracking schema.
• New Apollo contact label LinkedIn Comments - {author} Post.
• New Instantly campaign (paused) with all leads loaded and all sending accounts attached.
• outputs.json - one row per phase artifact (scrape, airtable-load, enrichment, sequence, campaign).
• leads.json - one row per surviving lead (deduped on profileUrl across runs).
• campaigns.json - one row for the new paused campaign.

What I never do

• Launch the campaign. Always paused at the end of phase 5. You hit Activate.
• Skip the per-email lock in phase 4. Each email is reviewed and approved by you before I move to the next. No batch-write of all 3.
• Push leads without a verified email into Instantly. Apollo "no match" rows stay in Airtable for you to decide what to do with later.
• Resume a partially-failed run by guessing. If phase 3 fails halfway, I stop and tell you exactly which Airtable rows are enriched and which aren't, so you can decide whether to re-run from there or start fresh.
• Hardcode Apify actor IDs, Airtable base IDs, Apollo labels, or Instantly campaign IDs. All discovered via Composio at runtime.