outreach-sprint
Skill1 starsChanged 17 days agoMITView on GitHub
Run a time-boxed outreach push with Claude as operator: daily math, warm rings, target research with agents, approval-gated sends, follow-up calendar, reply handling, scoreboard and retro. Use for \"run our outreach for the next two weeks\", \"get us our first customers this month\", or to keep a push already under way on track.
In this repository16 other files it gives its agents.
- brandingSkill
- business-model-canvasSkill
- customer-discoverySkill
- experimentsSkill
- facts-assumptions-unknownsSkill
- founder-stackSkill
- gtmSkill
- jtbdSkill
- learnSkill
- marketingSkill
- problem-discoverySkill
- salesSkill
- secondary-researchSkill
- startup-scorecardSkill
- value-proposition-canvasSkill
- where-to-playSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 11
Read discovery/00-CONTEXT.md, discovery/LEDGER.md and discovery/FIELD-NOTES.md if they exist.
- Distinctive: line 28
Every reply moves up one rung: Write the Advance before answering [Rackham].
- Distinctive: line 37
No invented facts in copy.
- Distinctive: line 38
No guessed emails, no scraped personal data beyond what the person published for work, and an opt-out line in cold email.
- Distinctive: line 40
Stop when the goal is met or the window closes, then run the retro.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Outreach sprint
The execution loop for a 1–4 week push. /sales and /marketing hold the method; this skill runs it day by day and turns the results into field notes. It was built from a real founder push, and each lesson it carries is tagged [Field].
Start here
- Read
discovery/00-CONTEXT.md,discovery/LEDGER.mdanddiscovery/FIELD-NOTES.mdif they exist. Your own past numbers beat any benchmark. - Get four things or ask for them: the goal as a number and a date, the offer and price, who signs, and which channels and accounts the founder can send from. If you can't ask, state assumptions (graded E0) and go on.
- Open
discovery/sprints/<date>-<name>.mdfrom TEMPLATES §1. Everything below is logged there, in the repo, the same day. Scratch folders get wiped.
The loop
Day 0: plan
- Daily math (
TEMPLATES.md§1 in this skill's folder, same formula as../sales/playbooks/first-customers-founder-led-sales.md): customers needed ÷ close rate ÷ reply-to-conversation rate ÷ reply rate per account = accounts to start, spread over the days before the last start date (deadline minus cadence length minus time to close). Take the reply rate from the founder's own sends orFIELD-NOTES.md. With no baseline, run the math at 2% and 10% and plan for the low end. If cold can't reach the goal by the date, say so now. The fix is more warm volume or a new date, never a lower price. - Rings before cold [Lavingia] [Fitzpatrick]. List 15–30 ring-1 names (people who trust the founder) and ring-2 doorways (cohort leads, community hosts, people who posted about the problem). Ring 1 is asked for intros first. Only the founder sends ring-1 messages, or approves each one by name.
- One ask per awareness stage (
../sales/playbooks/first-ask-design.md). Draft the ladder (reply → sample → low-stakes try → real use) and the gift; track versions in TEMPLATES §3. Founder approves the copy. - Plan every touch on day 0: 5–7 per account over about two weeks, each adding something new: touch 1 day 0; touch 2 on another channel day 2–3; touch 3 the gift or a new angle day 5; touch 4 a visual or proof day 7; touch 5 a new angle day 10; a polite close day 12–14 [Kazanjy] [Ross]. Put them on the sprint calendar. Start no new account after the last start date.
Daily: run
5. Targets via research agents (brief in TEMPLATES §2). Each agent gets an exclusion file, a JSON output path it rewrites every three entries, a time box, and "fewer is fine; never guess an address". Verify every person by searching name plus company. Never trust a profile URL an agent produced.
6. Send only with approval. The founder approves each message, or a named batch. Before each send, check recipient, subject and body. If a send looks like it failed, check Sent before resending. Tag every link with ?ref=<channel>-<variant>. Log every send.
7. Respect the channel's limits [Field]: personal addresses only; a new domain stays at tens of cold emails a day across all its mailboxes; bounces under ~3%. Put volume where the person answers (often their social network, not email). For launch sites, directories and communities, pick 3–5 from ../marketing/reference/places-to-post.md, each with its own ?ref.
8. Check replies twice a day (morning, late afternoon) and answer within two hours using the reply scripts (TEMPLATES §4). Hourly polling finds nothing and burns attention [Field].
9. Every reply moves up one rung. Write the Advance before answering [Rackham].
Weekly: score and decide
10. Scoreboard by ref (TEMPLATES §5): touches, replies from buyers, Advances, calls, trials, paid. Accepts, opens and impressions go in a side column and never drive decisions [Fitzpatrick] [Ries].
11. Judge a channel only after its window: 5–7 touches per account over one to two weeks for outreach [Kazanjy] [Ross], about a month for content and communities [Weinberg]. Keep the channel, change one variable, or kill it.
12. Retro into field notes (TEMPLATES §6). One line per lesson, with n, date and grade, in discovery/FIELD-NOTES.md. Add assumptions to the ledger through /founder-stack.
Hard rules
- Nothing goes out without the founder's approval: no message, post, comment or connection request. Logins, payments and CAPTCHAs are always handed back to the founder.
- No invented facts in copy. Every number or news hook needs a source the reader could check.
- No guessed emails, no scraped personal data beyond what the person published for work, and an opt-out line in cold email.
- One variable changed per test. Compare variants on equal list quality.
- Stop when the goal is met or the window closes, then run the retro.
Output
Each day: the sprint log, sends and replies, and the next day's list for approval. Each week: the scoreboard, keep/change/kill per channel, and the field-notes and ledger delta. End with: where the goal stands, the riskiest assumption and tomorrow's first action.
Hand-offs
- Method:
/salesplaybooksfirst-customers-founder-led-sales,first-ask-design,cold-outreach-and-cadence,/marketing(founder-posts-and-communities, channel-selection). - Tests and grading:
/experiments. Ledger:/founder-stack.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

