agentleFS
marichalnickmarichalnick/sales-skills-for-claude/skills/signal-verified-outreach/SKILL.md

signal-verified-outreach

Skill0 starsChanged 29 days agoMITView on GitHub

Drafts cold outreach (email or LinkedIn) from a verified trigger. Checks every claimed fact against a primary source before it reaches the copy, picks the right stakeholder, avoids repeating earlier touches, and writes one short message with one question and one low-friction call to action. Use when the user asks to write a cold email, prospecting message, outreach for a company or person, "who should I contact at X", or a batch of outreach for a list of accounts.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 10 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Look for dated, specific events from the last ~90 days:

    line 22
  2. Distinctive:

    Check the specific claim, not just the topic.

    line 60
  3. Distinctive:

    If nothing survives verification, say so.

    line 66
  4. Distinctive:

    Ask: who feels the pain, who owns the risk, who controls budget, who evaluates technically, who could block?

    line 70
  5. Distinctive:

    Prefer someone not contacted recently.

    line 71
  6. Distinctive:

    If the user has relationship context (a past call, an intro offer), that outranks everything else.

    line 72
  7. Very specific:

    Follow references/style-guide.md.

    line 83
  8. Distinctive:

    Every follow-up adds something new: a different angle, a relevant resource, a new signal.

    line 90
  9. Distinctive:

    No guilt-trip break-up lines, no fake urgency.

    line 93
  10. Very specific:

    If the user works across sessions, also save a short ACCOUNT-<name>.md with the verified triggers, sources and chosen stakeholder, so …

    line 121
  11. Distinctive:

    If several accounts were requested, start with a short ranked table (account, trigger, age, person, confidence) and list the ones you …

    line 123
  12. Distinctive:

    If verification knocks out the trigger a message was built on, tell the user plainly what broke and rebuild from what is verified, or …

    line 126

≈ “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

Signal-Verified Outreach

Turn "write outreach to this company" into a message that is true, timely and relevant, the way a careful SDR would do it with unlimited time to check every fact.

The one rule: verify before you hook

AI-assisted outreach fails in a specific way: it invents a convincing trigger. A regulation that doesn't exist, a date nobody announced, a press release paraphrased backwards. One fabricated fact in a message to a senior buyer costs more credibility than a hundred good messages earn.

So: no claim reaches the copy unless you opened the primary source and read the sentence it rests on. A label like "verified" in a spreadsheet or CRM is not a substitute.

Workflow

1. Understand the ask

Get: target company (and person, if given), what the user sells, and any prior contact. If the offer or ICP is unclear, ask once, briefly.

2. Find candidate triggers

Look for dated, specific events from the last ~90 days:

TriggerWhy it matters
Leadership change (new CISO, CRO, CMO…)First 90 days = priorities are being reset
Funding, acquisition, expansionBudget and pressure to scale
Relevant hiringOwnership and budget are being created
Product or technology launch / rolloutNew risks and new processes
Regulation that applies to themPressure, but not proof of intent
Something the person posted or saidThe strongest signal: their own words
Tech change (new tool adopted, vendor dropped)A process is being rebuilt
Public frustration with a current vendor (reviews, posts)Openness to alternatives

A standing fact ("they are a big bank") is context, not a trigger.

Timing windows (rules of thumb, tune them with your own data):

  • New leader: roughly weeks 2 to 8 in the role. Too early they are onboarding, too late priorities are set.
  • Funding or acquisition: the first few weeks after the announcement.
  • Job posting: while the role is open, and again shortly after someone is hired.
  • Anything older than ~90 days is context, not a trigger.

Stacked signals beat single signals. Two independent, verified signals on the same account (for example a new CISO and a relevant job posting) are a much stronger reason to reach out than one. Rank accounts with stacked signals first, and say which signals stack.

3. Verify each trigger

Source hierarchy

LevelSourceUse in copy?
1Company's own release, filing, careers pageYes
2Regulator or government sourceYes
3The person's own post or statementYes, quote carefully
4Reputable journalismWith care; prefer finding level 1 to 3
5Databases, aggregatorsResearch only
6Search snippets, blogs, SEO contentNever stated as fact

For every trigger you plan to use:

  • Open the primary source: the company's own site or press release, the regulator's own page, the person's own post.
  • Check the specific claim, not just the topic. Read the surrounding sentence.
  • Watch for two failure modes:
    • The claim does not exist. A secondary source added a date, number or name that the primary source does not contain.
    • The claim is backwards. "We solved X" paraphrased as "they struggle with X".
  • Note the age. Signals decay fast: weeks, not months.

If nothing survives verification, say so. Do not patch a broken trigger. A real but older fact, framed honestly, beats a fake new one.

4. Pick the person

Choose the role that owns the problem the trigger creates, not the highest title.

  • Ask: who feels the pain, who owns the risk, who controls budget, who evaluates technically, who could block?
  • Prefer someone not contacted recently. Never repeat an angle already used with the same person.
  • If the user has relationship context (a past call, an intro offer), that outranks everything else.

5. Reconcile with prior touches

What was already said and asked? If there is nothing genuinely new, recommend waiting instead of sending a reworded repeat.

6. Match the message to the level

  • Executives (above the line): business outcome, risk, timing. Shorter. One question about priorities.
  • Practitioners (below the line): the day-to-day problem, the workflow, the tool. One question about how they handle it today. Same trigger, different angle. If you contact both at one account, never send them the same message.

7. Write the message

Follow references/style-guide.md. In short:

  • 50 to 120 words. One idea. One question. One call to action.
  • Open with them (the trigger), not with you or your company.
  • State the hypothesis as a question, never as a fact about their situation.
  • Low-friction CTA: a concrete slot or calendar link beats "can I have your number?".

8. Follow-ups (only if asked)

  • Every follow-up adds something new: a different angle, a relevant resource, a new signal. Never "just bumping this".
  • Typical rhythm: day 3 to 4, day 8 to 10, then a short, honest closing note ("I'll stop here; if timing changes, happy to talk").
  • Switch channel once (email → LinkedIn or phone) before switching person.
  • No guilt-trip break-up lines, no fake urgency.

9. Human approval gate

Before anything is sent, the user must check: names and titles, every factual claim, no unsupported social proof, the CTA, and whether the tone fits the relationship. Present the draft for approval; never imply it has been or will be sent automatically. Respect opt-outs: anyone who asked not to be contacted is excluded.

10. Output

For each message, first an evidence table:

TypeClaimSource (level)DateUse
Fact…URL (1)…Safe to mention
Inference…derivedAsk as a question
Unknown…noneDiscovery gap

Then:

Account / person / role
Trigger: <what happened> — <primary source URL> — <age>
Why this person: <one line>
Hypothesis: <the pain you think this creates>

<subject line, if email>
<message>

Confidence: high / medium / low, and why

If the user works across sessions, also save a short ACCOUNT-<name>.md with the verified triggers, sources and chosen stakeholder, so discovery-call-prep and call-to-crm can build on it instead of re-researching.

If several accounts were requested, start with a short ranked table (account, trigger, age, person, confidence) and list the ones you dropped with the reason.

When something doesn't add up

If verification knocks out the trigger a message was built on, tell the user plainly what broke and rebuild from what is verified, or recommend not sending. Fewer true messages beat more false ones.

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.