agentleFS
vivekkhimanivivekkhimani/gtm-tools-template/.agents/skills/email-personalization/SKILL.md

email-personalization

Skill1 starsChanged 5 months agoMITView on GitHub

Draft a 90-word personalized first-touch cold email for a [COMPANY] prospect. Takes one company row (from leads_scored.csv or leads_enriched.csv) plus an optional buying signal, and produces a draft that follows the v1 sequence pattern. Use when drafting Email 1 of the 4-touch sequence (outreach/sequence-v1.md). Also triggers on "draft outreach email", "write cold email", "personalize email for {company}".

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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
  1. Distinctive:

    NOT for follow-up emails (2, 3, 4) — those are thread replies with simpler templates already in the sequence doc.

    line 14
  2. Distinctive:

    DO NOT make up a signal. If you don't have one, return an explicit "needs signal research" output instead of a fabricated email.

    line 39
  3. Very specific:

    Draft: Follow the structure in outreach/sequence-v1.md Email 1.

    line 41
  4. Distinctive:

    Use specific [INDUSTRY] terms when they fit.

    line 92

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Email personalization

Generate a personalized Email-1 cold outreach draft for a [COMPANY] ICP target, matching the v1 sequence pattern in outreach/sequence-v1.md.

When to use

  • Drafting first-touch emails for the top tier (T1/T2) of pipeline/leads/leads_scored.csv.
  • Re-drafting after a signal changes (new buying signal, leadership change, compliance event).
  • NOT for follow-up emails (2, 3, 4) — those are thread replies with simpler templates already in the sequence doc.

Inputs

InputRequiredSource
Company rowYesA single row from pipeline/leads/leads_scored.csv or leads_enriched.csv. At minimum: Company Name, Short Description, Keywords, Company Country, # Employees.
Recipient name + titleYesFirst name, title, ideally verified via Findymail/Hunter
Buying signalOptional but recommendedOne-sentence specific fact: <!-- TODO: 2–3 example signals for YOUR industry, e.g. "took delivery of new capacity in March 2026", "hired a [BUYER_TITLE] in February", "recent compliance finding in [GEO_PRIMARY]" -->
AngleOptionalOne of: [ANGLE_A], [ANGLE_B], [ANGLE_C]. Default: pick the one best-matched to the signal. <!-- TODO: name your 2–3 messaging angles -->

Process

  1. Read positioning/messaging.md, positioning/ICP.md, and outreach/sequence-v1.md. The skill must conform to the voice and structural conventions already defined there. Do not invent claims that aren't in messaging.md.
  2. Classify segment for the company (A/B/C per ICP.md). If unclear, infer from Short Description + Keywords, but prefer to ASK rather than guess on segment.
  3. Choose angle. Pick the angle that maps to:
    • The provided signal. <!-- TODO: map signal types → angles, e.g. a compliance event → [ANGLE_A] -->
    • The segment (Segment A favors [ANGLE_A]; Segment B favors [ANGLE_B]; Segment C favors [ANGLE_C]).
  4. Generate signal sentence. This is the most important sentence — first line of the email. It must reference a specific, verifiable fact about the company. Source it from:
    • The provided signal input (preferred).
    • The Short Description column (less specific but better than nothing).
    • A recent news snippet if web search is available.
    • DO NOT make up a signal. If you don't have one, return an explicit "needs signal research" output instead of a fabricated email.
  5. Draft. Follow the structure in outreach/sequence-v1.md Email 1. Constraints:
    • 80–95 words in the body.
    • First sentence is the signal reference + the pain hook.
    • One proof point (citing the canonical traction: "[METRIC_DEPLOYMENTS] / [METRIC_USERS]" — or "replaced [INCUMBENT] at [ANCHOR_CUSTOMER]" if a switch-out proof lands harder for the segment).
    • One soft CTA ("worth a 15-min call?").
    • Plain text, no HTML, no formatting.
    • Lowercase, no exclamation marks, no "Hope this finds you well" or similar dead opens.
  6. Subject line. Generate 3 candidate subject lines following the patterns in outreach/sequence-v1.md. ≤8 words. Lowercase. No question marks at the end unless it's a real question. Reference a specific token (a product/category term, company name, role) — don't go generic.
  7. Output as YAML so it slots cleanly into Smartlead/Instantly imports:
to: {recipient_email}
recipient_name: {FirstName} {LastName}
recipient_title: {Title}
company: {Company}
segment: {A|B|C}
angle: {[ANGLE_A]|[ANGLE_B]|[ANGLE_C]}
signal_used: "{one-sentence signal}"
subject_options:
  - {subject 1}
  - {subject 2}
  - {subject 3}
body: |
  Hi {FirstName},

  {body 80–95 words}

  — [FOUNDER]
  [DOMAIN] · [CALENDLY]
qc_checks:
  word_count: {N}
  has_signal: true|false
  has_one_cta: true|false
  no_banned_phrases: true|false  # checks "hope this finds you", "circling back", "just checking in", "synergy", "leverage", "solution", "platform"

Banned phrases (never use)

  • "Hope this finds you well"
  • "Just checking in"
  • "Circling back"
  • "Touching base"
  • "Synergy", "leverage", "solution" (as nouns), "platform" (generic)
  • "10x", "best-in-class", "world-class"
  • "I noticed you on LinkedIn..." (we are not on LinkedIn for this audience)
  • Any subject line that's only "{Company}" with no other context

Voice constraints

  • Use specific [INDUSTRY] terms when they fit. <!-- TODO: list the concrete vocabulary your buyer uses — product/category terms, regulation/standard codes, equipment, certifications — so drafts sound native, not like generic SaaS copy. -->
  • The reader is a [INDUSTRY] operator, not a SaaS buyer. Match their register.
  • Honest claims only — canonical traction is [METRIC_DEPLOYMENTS], [METRIC_USERS], [METRIC_COMPLETION], anchor customer [ANCHOR_CUSTOMER] with the incumbent replaced (name them only if cleared for external use). Don't inflate beyond these. Keep the wedge framing in cold email (the budgeted entry pain) — do NOT cold-open with "[PRODUCT_ONELINER]"; "platform" stays a banned generic noun here.

Failure modes (return explicitly rather than fake it)

  • No signal available: return status: needs-signal-research with a list of 3–5 specific signal queries to run. <!-- TODO: example queries for YOUR industry's data sources -->
  • Segment ambiguous: return status: needs-segment-decision with two interpretations and your recommendation.
  • Recipient title doesn't match [COMPANY]'s buyer personas: return status: wrong-persona with a recommended persona to find instead.

Example invocation

Draft an Email-1 for [TARGET_COMPANY] (row in leads_scored.csv) targeting their [BUYER_TITLE]. Signal: <!-- TODO: a specific, verifiable fact, e.g. "they were quoted in [TRADE_PRESS] last month about [a pain in your space]". -->

Expected output: a YAML block as specified above, with the angle matched to the signal, 3 subject candidates, and a body that opens with the signal reference.

Integration

This skill is invoked manually per email during the first weeks (we want human review of every draft). Once you've shipped ~50 drafts and feel confident the voice is right, wrap it in scripts/draft_emails.py to batch-process the top 20 weekly.

Track each draft in outreach/drafts/_template/{company-slug}.yaml (copy _TEMPLATE.yaml per prospect). Drafts are .gitignored if they include recipient emails; the templates without recipient emails are fine to commit for voice tuning.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.