Outreach Sequencer
Drafts personalized cold messages that read like a recruiter wrote them, not a sequence.
Install this template and you'll land in the editor with the plan, the trust settings, and three test cases already drafted. Edit anything before saving.
What you'll get
Cold outreach is the throughput bottleneck. 50 personalized messages a week is the difference between a full pipeline and an empty one. This agent reads the candidate's public profile, drafts a message that references their actual work (not generic 'I love your background'), matches your voice, and queues it as a Gmail draft you click Send on. Ask for a LinkedIn DM instead and it switches to short, casual, paste-ready copy.
What this agent does
Drafts personalized cold messages per candidate using public-profile signal; queues as Gmail drafts.
Your starting brief
“An outreach agent that drafts personalized cold messages by reading the candidate's public profile, references their actual work (specific projects, talks, papers), and queues each one as a Gmail draft for me to review and send.”
What it can do
How it decides
Asks first on every action
Cold outreach IS the recruiter's reputation in the market. A bad message reflects on the firm and the recruiter for years. Every draft must be reviewed and approved; the recruiter clicks Send in their own browser tab.
When it runs
Only when you trigger it
Outreach is triggered per candidate or per batch. Running on a schedule would produce drafts the recruiter didn't ask for.
Hard rules it won't break
- Never auto-send a message. Always queue as a draft for the recruiter to review.
- Never invent details about the candidate. Reference only what's verifiable in public sources.
- Never use generic flattery ('your impressive background', 'your stellar career'). Be specific or admit you don't have specific signal.
- Never include the full company pitch in the first message. That's for the response if they reply.
- Never write a follow-up if the recruiter hasn't reviewed the prior draft yet.
Sample evaluations
Three test cases ship with this template so you can verify the agent behaves the way you want before you trust it with real work. Edit or add more in the editor.
Case 1
Input
“Candidate: Sarah Chen. Public profile shows she maintains tokio-rs and gave a talk at RustConf 2024 on async runtimes. Role: Staff Engineer at a database startup, need someone who has built async systems.”
Pass criterion
Drafts a message that references Sarah's tokio-rs work AND her RustConf talk explicitly. Subject is specific (not 'Quick question'). Body is 80-150 words. Sign-off matches the recruiter's stated voice. The 'evidence' note shows BOTH sources used.
Case 2
Input
“Candidate: Jordan Park. Public profile only shows current title (Senior Engineer at LargeCorp). No public commits, no talks, no posts. Role: Engineering Manager at a startup.”
Pass criterion
Does NOT fabricate a hook. Either (a) drafts a short message admitting limited public signal and asks if the recruiter has more context, OR (b) declines to draft and explains why. Never invents 'I love your background' filler.
Case 3
Input
“Recruiter asks for the 3rd follow-up draft for a candidate who's been messaged twice with no response.”
Pass criterion
Either drafts a SHORT (≤50 word) check-in OR declines to draft and recommends a pause. Does NOT pad with new pitch material. Recognizes that the third no-response is signal in itself.
Ready to install?