outreach-playbook
Skill1 starsChanged 47 days agoMITView on GitHub
Draft LinkedIn outreach (hiring managers, recruiters, referrers) from the candidate's message playbook, gate every send on approval, and track sends/follow-ups with the toolbelt.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- apply-in-browserSkill
- find-jobsSkill
- review-loopSkill
- session-startSkill
- tailor-applicationSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 57
Plain register, zero pressure: No urgency ("urgent", "ASAP"), no generic flattery, no "kindly", no exclamation stacks.
- Very specific: line 84
Read the playbook file, pick the variant, gather candidate facts (dossier, prefs, answers) and target facts (job show / app show, the …
- Distinctive: line 102
After the candidate approves and sends, log it so the follow-up rule survives the session:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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 messages (the playbook)
The candidate's formula bank lives at workspace/profile/outreach-playbook.md
(families A–F with the exact templates). Read it before drafting. If it is
missing, ask the candidate for it; draft from the matrix below meanwhile and
save the result there. It is personal data (workspace/ is gitignored) — never
commit it, never paste it into logs or third-party services.
Author-facts guard. The playbook was supplied by the candidate but written
for someone else: its filled-in proof points and sign-off are the original
author's examples. Keep the formulas; fill every bracket and every quantified
claim from the candidate's own dossier (dossier search), answers, prefs,
and the job/app records. If the dossier doesn't support a number the template
wants, leave the bracket and ask — never copy the author's figures and never
invent. If the playbook genuinely belongs to the candidate, they'll say so; then
and only then may its numbers be used.
The 6 rules (every message, every channel)
- Under 100 words — long messages get skimmed and skipped.
- Personalize the first line (their team, product, or a post). Generic = deleted.
- Lead with value/fit, not a request. Never open with "Are you hiring?"
- Anchor ONE concrete, quantified proof point — traceable to the dossier.
- End with ONE low-friction ask (a 15-min chat, a pointer, a yes/no).
- Replies refund InMail credits — target hard, and follow up once after 4–5 days (tracked; see below).
Scam-aware drafting (India — every referral ask, invite, and follow-up)
Indian professionals are hyper-alert to job scams (fake recruiters, phishing, "refer me" spam). Anything that pattern-matches a scam gets ignored or reported — so every draft gets this check before the humanizer pass:
- Never ask for "insider" anything. No "insider read", "insider info", "what's it really like inside". The ask is a referral or a pointer, never information extraction — asking for confidential-sounding info makes the recipient uncomfortable and scam-alarmed.
- Lead with the verifiable thread. Name the real connection in line 1 — mutuals in common, ex-employer overlap, college — only what both LinkedIn profiles actually show. A checkable thread is the strongest "real person" signal; vague or faked familiarity is the strongest scam signal. Never invent or stretch a tie.
- Name the exact req. Role title (+ req ID when known) and location in the first or second line. "Roles at your company" reads as mass spam.
- Give the referrer a forwardable fit line. Level, years, stack, one quantified proof point — what they need to justify a referral in 10 seconds.
- The ask is a portal action, not a conversation. "Would you be comfortable referring me? Happy to send my resume" — offer the resume yourself; never ask them to click off-platform links or send anything to an email address (that is phishing shape), and never ask for their details.
- Plain register, zero pressure. No urgency ("urgent", "ASAP"), no generic flattery, no "kindly", no exclamation stacks; one sincere out ("completely understand if not") and no follow-up nagging beyond the single tracked E-follow-up.
Scam-avoidance and conversion are the same move: specific, verifiable, easy to act on, zero obligation.
Pick the variant
| Situation | Variant |
| Open role → hiring manager | A1 (direct fit) |
| No open role, admire the team | A2 |
| They posted / spoke publicly | A3 |
| In-house recruiter, reach or reply | B1 |
| Cold outreach to an agency recruiter | B2 |
| One-line pitch inside a recruiter InMail reply | B3 |
| 2nd-degree connection at the target company | C1 |
| Former colleague at the target company | C2 |
| Ask someone in your network for an intro | C3 (include the 2-line blurb) |
| Connection request note — free, no InMail | D1 manager / D2 recruiter / D3 peer (≤ 300 chars — count them) |
| Follow-up, once, 4–5 days later | E1 gentle nudge / E2 new proof point |
| InMail subject line | F bank |
Workflow
- Read the playbook file, pick the variant, gather candidate facts (dossier,
prefs, answers) and target facts (
job show/app show, the person's profile or post — read it in the browser, don't guess it). - Draft into the application's
notes.mdartifact (or a scratch markdown file when outreach precedes an application). Check the caps: < 100 words for messages; connection notes ≤ 300 chars on free accounts, up to ~500 on paid plans — read prefslinkedinplanfor the candidate's plan and verify the live counter in the compose UI (LinkedIn adjusts limits). One ask per message. Check every draft against the scam-aware rules above (especially: no "insider" asks). Then run every draft through thehumanizerskill before it is shown: the playbook formulas are skeletons, and the humanizer pass strips the AI tells so the message reads like the candidate wrote it. Facts and numbers stay exactly as gathered in step 1. - Show every draft and wait for an explicit yes — nothing leaves the machine without it (prime directive 1). A few at a time, never bulk blasts; personalized per person, not template spam.
- After the candidate approves and sends, log it so the follow-up rule survives
the session:
node bin/applymate.ts outreach log --target "<name>" --role hiring-manager|recruiter|referrer --channel inmail|connection-note|dm|email --variant A1 [--app <appId>] [--company <co>] [--note "..."] - Chase on schedule, not from memory:
node bin/applymate.ts outreach list --dueat the start of any outreach session. Send the E-variant follow-up ONCE on approval, thenoutreach followup <id>. When a reply lands, record it:outreach replied <id>— the toolbelt then stops listing it as due.
$ARGUMENTS
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.

