Outreach Batch

Outreach Batch

You are managing a maker outreach campaign for Your Project. Your goal is to get tool makers to claim their listings, increasing engagement and moving toward revenue.

Before Starting

Check for context first:

• How many emails have already been sent this session? (Check playbook)
• Did previous batches get any claims? (Check magicclaimtokens on prod)
• Has S&QA approved more outreach, or said to wait for signal?

Context Needed

1. How many targets? (Default: 15-20 per batch)
2. Star threshold? (Default: 1000+ GitHub stars)
3. Any tools to exclude? (Already emailed list)

How This Skill Works

Mode 1: Full Campaign (first time)

Run all 6 steps below in order. Takes ~30 minutes with departments.

Mode 2: Next Batch (adding more)

Skip to Step 1 with exclusions from previous batches. Reuse existing templates.

Mode 3: Follow-Up

Check claim analytics, resend to non-responders with adjusted angle after 7 days.

The 6-Step Process

Step 1: Find Targets (Backend)

SELECT slug, name, github_stars, url FROM tools
WHERE status='approved' AND maker_id IS NULL
AND github_stars > {threshold}
AND slug NOT IN ({already_emailed})
ORDER BY github_stars DESC LIMIT {count}

Step 2: Find Emails (Backend)

For each target, try in order:

1. gh api users/{owner} --jq '.email'
2. gh api repos/{owner}/{repo} --jq '.owner.email'
3. Check CONTRIBUTING.md, README security contacts
4. Check package.json/composer.json author fields

Expected hit rate: ~70%. Skip tools with no public email.

Step 3: Audit Target Pages (MCP Dept)

BEFORE sending, verify each tool page has:

• [ ] Correct install_command (NOT wrong tool — Cal.com had mailhog!)
• [ ] Non-truncated tagline
• [ ] Description that's more than just the tagline
• [ ] Correct category

Fix issues via Backend before proceeding.

Step 4: Write Emails (Content)

Two template types:

• Has MCP views (>0): "agents recommended your tool X times"
• Zero views: "AI agents couldn't find your tool — we fixed that"

Rules: under 80 words, from "Pat", include your-project.ai/claim/magic?tool={slug}, no upsell.

Step 5: S&QA Review

Send plan to Strategy before dispatching emails. S&QA checks:

• Are we sending too many too fast?
• Did we audit the target pages?
• Is the timing right (wait for signal from previous batch?)

Step 6: Send (Master via SSH)

await send_email(to=email, subject=subject, html_body=body)

Via ~/.fly/bin/fly ssh console -a your-project. Max 5 per call. 1s delay between.

Proactive Triggers

Surface these WITHOUT being asked:

• Broken claim links: If /claim/magic route changes, ALL email links break. Test before every batch.
• Wrong install commands: If a tool's install_command is wrong, the maker's first impression is "these people don't know my tool." Always audit.
• Diminishing returns: If batch 1 got 0 claims after 48 hours, don't send batch 3. Investigate the pitch first.
• S&QA override: If S&QA says "wait for signal," respect it. More outreach ≠ better outreach.

Output Artifacts

| Request | Deliverable | |---------|------------| | "Run outreach batch" | 15-20 personalized emails sent, /tmp/claimemailsbatch_N.md | | "Find targets" | Slug + name + stars + email list | | "Check responses" | Claim analytics from production DB | | "Follow up" | Re-send with adjusted templates after 7 days |

Gotchas (from real experience)

• Cal.com had mailhog install command — ALWAYS audit pages
• Claim links used ?tool=slug which was broken — fixed with fallback on 2026-03-30
• Apache mailing lists (dev@airflow.apache.org) are not appropriate for cold outreach
• syncthing only has security@ address — use cautiously
• Em dashes in email HTML cause Python SyntaxError via SSH — use -- instead

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Pattyboi101
• Source: Pattyboi101/oats-autonomous-agents
• License: MIT

Install and usage instructions live in the source repository linked above.