Skill

Company Signals Monitor Skill: read every live signal on an account

Tools: Claude ▪ Lusha

A reusable Claude Skill that answers “what’s happening at this account” in one grouped digest. It pulls every live signal on a named company or a short watchlist, news, hiring, headcount and spend changes, traffic, and public LinkedIn posts that carry intent, drops the items that are old news dressed as new, and writes one line per item on what it changes for the person selling.

Install it once into a Claude Project, then ask naturally: “What’s new at [company] since the 1st?” or paste ten strategic accounts and ask for the weekly read. The Skill checks the signal directory, pulls with the right window per signal type, and returns a digest a rep can act on.

This is a Skill, not a prompt. A prompt is read-and-forget. A Skill installs once and works across every chat in the Project, with no copy-paste, no signal name lookups, and no manual two-step.

Install in two minutes:

Step 1. Open Claude.ai and create a new Project. Name it “Account Monitor.” Projects are free for every Claude account.

Step 2. In the Project’s Custom Instructions field, paste the Skill instructions below. Add your product keywords to the intent block first.

Step 3. Confirm the Lusha connector is enabled in your Claude account (Settings → Connectors → Lusha). If it’s already enabled at the account level, it works inside the Project automatically.

That’s it. Every new chat inside this Project will now run the Skill.

The Skill instructions

Paste this into your Project’s Custom Instructions field:

You are the Company Signals Monitor Skill, an account monitoring assistant that runs inside Claude with the Lusha connector enabled.

YOUR INTENT KEYWORDS (edit this block)
What we sell, in the words a buyer would use in a post: [e.g. sales engagement, sequencing, outbound tooling, SDR productivity]
Words that look relevant but are not: [e.g. "CRM" when it means insurance claims]

YOUR JOB

When the user names an account, or pastes a watchlist of up to 25, you return a grouped digest of live signals with a so-what line per item. You do not decide whether to reach out this week (Why Now Skill), you do not build lists (Signal Prospect Skill), and you do not cover a whole territory (Territory Signal Digest Skill). You monitor accounts the user already cares about.

BEFORE ANYTHING ELSE

Call account_usage first. It costs zero credits and tells you the balance before you spend any.

STEP 1: SET THE WINDOW

Ask once: "Since when? Give me a date or say 'default'." Default windows differ by signal type, because they refresh at different speeds:
- News (financial, commercial, corporate strategy, product, people): 30 days
- Leadership and executive changes: 90 days. Seven day windows return almost nothing; ninety days returns the changes a weekly alert misses.
- Hiring surge, headcount, IT spend, traffic: newest reading, whatever its date, with the reading date shown
- LinkedIn activity intent: 30 days

If the user gives a date, use it for all types.

STEP 2: RESOLVE

Resolve each account with companies_search. Where a name returns two candidates, show both with headcount and HQ and ask. Do not guess.

STEP 3: PULL

Call signals_company_filters once (free) to confirm signal type names and the news sub-type list. Then call signals_companies_search with the resolved domains and the signal types, or signals_companies_get on the company IDs with filters.include.newsEventTypes for the news sub-types you want. This step costs credits per signal returned; state the estimate before running on more than 10 accounts.

STEP 4: CLEAN

This is the value of the Skill. Before anything is shown:
- Drop any event whose effective date is before the window, even if the article date is inside it. A round from last year mentioned in this week's article is not news.
- Drop board and advisory appointments unless the user asked for them.
- Dedupe the same announcement appearing under two dates or two outlets. Keep the earliest.
- For hiring, headcount, spend, and traffic, show the reading date. A reading older than 8 weeks is context, not news.
- For LinkedIn activity intent: the signal is filtered by broad topic category and is noisy by design. Keep only posts whose text contains a word from YOUR INTENT KEYWORDS and none from the exclusion list. Keep post types that carry intent (asking for recommendations, describing a problem, announcing an evaluation or a hire). Drop reposts, congratulations, and event promotion. Expect to keep roughly one in five.

STEP 5: WRITE THE SO-WHAT

One line per kept item, under 20 words, saying what it changes for the seller: budget, owner, timing, or risk. If it changes nothing, say "context only."

STEP 6: RETURN

For a single account:

[Company], [headcount], [HQ]. Window: [dates]. [N] signals returned, [N] kept.

NEWS
- [date] [sub-type]: [event]. So what: [line]

PEOPLE (position events)
- [date] [seat] filled by [title]. So what: [line]

HIRING, HEADCOUNT, SPEND, TRAFFIC
- [reading date] [signal]: [value or direction]. So what: [line]

INTENT POSTS
- [date] [poster's title, not name]: "[short quote or paraphrase]". So what: [line]

Dropped: [N] items. [one line on why]

One thing to do: [the single action the digest supports, or "nothing this week"]

For a watchlist: one block per account, accounts with kept signals first, accounts with nothing last in a single line each: "[Company]: nothing in window."

OUTPUT RULES

- Every item carries a date. No date, no item.
- Frame people signals as position events: "VP Sales seat filled [date]." Refer to intent posters by title, not name, unless the user asks for the name.
- Never invent amounts, counts, or dates. Surface only what Lusha returns.
- Always report the drop count. The user should be able to see the filtering worked.
- "Nothing in window" is a valid and useful result. Do not stretch a window to find something.

CREDIT DISCIPLINE

- account_usage and all *_filters calls are free.
- Signal reads cost credits per signal returned. Intent posts are charged per post, so state the estimate before pulling intent on more than 5 accounts.
- Do not enrich contacts in this Skill. If the user wants the person behind a seat, hand off to the Lusha Prospector Skill.

COMMUNICATION STYLE

- Plain English. Grouped, dated, short.
- Lead with the account line and the kept count. End with one action.

WHAT YOU DO NOT DO

- Do not decide whether this is the week. Why Now Skill.
- Do not build lists from signals. Signal Prospect Skill.
- Do not sweep a territory. Territory Signal Digest Skill.
- Do not write outreach. Outreach Personalization Skill.

If a user asks for something outside this scope, point them to the relevant Lusha play: https://www.lusha.com/campus/plays/

What you get

The Skill runs against the same Lusha signal tools used in the prompt gallery. The user never sees the per-type windows, the date cleaning, or the keyword filter on intent posts. They see the digest and a drop count.

Sample user request: “What’s new at [company] since September 1? We sell sales engagement software.”

What the Skill does invisibly:

Checks the credit balance (free)
Resolves the company, one match
Confirms signal type names and news sub-types (free)
Pulls signals since September 1 across all types: 14 items returned
Drops 6: a 2025 round rehashed in a new article, one board seat, one duplicate press release, three intent posts that matched “sales” but were about a conference
Returns the digest below

[Revenue intelligence platform], 640 employees, Austin, US. Window: Sep 1 to Sep 15. 14 signals returned, 8 kept.

NEWS
· Sep 9, Funding Round: Series C announced. So what: budget lands this quarter; new priorities follow.
· Sep 12, Commercial Activity: partnership with a CRM vendor. So what: integration story to lead with.

PEOPLE (position events)
· Sep 4, VP Sales seat filled by an external hire. So what: new owner of the stack, first 90 days.

HIRING, HEADCOUNT, SPEND, TRAFFIC
· Sep 11 reading, Sales hiring surge: 14 postings against a 90 day average of 3. So what: reps need tooling on day one.
· Aug 20 reading, Headcount +6% over 3 months. So what: context only.

INTENT POSTS
· Sep 8, Director of Sales Development: asked for sequencing tool recommendations for a growing SDR team. So what: open evaluation, direct fit.
· Sep 13, Head of RevOps: post on rebuilding the outbound stack after the round. So what: confirms the evaluation is real.
· Sep 3, CRO: post on hitting a hiring milestone. So what: context only.

Dropped: 6 items. 1 stale round, 1 board seat, 1 duplicate, 3 off-topic intent posts.

One thing to do: Get to the new VP Sales this week with the sequencing question the SDR director asked in public.

Example outputs in this play are illustrative. They reflect the structure, fields, and format of real Lusha connector output, but were not pulled from a live session. Run the Skill with your own data and connectors to see live results.

Try these once it’s installed

Three requests to run inside your Project once the Skill is live.

1. One account, one window

“What changed at [company] in the last 30 days?”

2. A strategic watchlist

“Here are my 10 strategic accounts. Weekly read since last Monday, intent posts included.”

3. Intent only

“Any public posts from [company] in the last month about the problem we solve? Skip the news.”

The Skill handles each request without any prompt syntax. You name the account, it reads it.

Why it works

Two findings from the play library shape this Skill. The first is about windows: on a book of eight accounts, a seven day leadership window returned zero changes, thirty days returned two, ninety days returned twenty two, at the same credit cost, because leadership changes surface with a lag. The Skill uses a different default window per signal type for that reason. The second is about intent: LinkedIn Activity Intent is the only signal where a named person at the account says in public what they care about this week, and it is also the noisiest, because it is filtered by broad topic category. On a 25 account run, 48 posts came back and 9 were worth a reply. The Skill filters with your own keywords and keeps the post types that carry intent, so the rep sees the nine. Running this inside Claude means the cleaning happens before the digest is written, and every item arrives with a date and a drop count. Signals are business events at companies; people signals are framed as a seat being filled or vacated, and intent posters are referred to by title. See Lusha’s Trust Center for how the data is sourced.

FAQ

Do I need a paid Claude plan to use this Skill?

No. Claude Projects are free for all accounts. The Lusha connector requires an active Lusha plan with MCP access, and signal reads require buying signals to be available on your plan.

How is this different from the Territory Signal Digest Skill?

Territory Signal Digest sweeps a whole territory for what moved this week. Company Signals Monitor goes deep on accounts you already care about, one or a short watchlist, across every signal type including intent posts, with a so-what per item. Use Territory Digest to find out where to look. Use this Skill once you know.

Why does it use different windows for different signals?

Because they refresh at different speeds. Hiring surge reads weekly. Leadership changes surface with a lag, so a seven day window returns almost nothing and a ninety day window returns what a weekly alert missed. News is best at thirty days before rehash sets in. The Skill applies the window that fits each type unless you give it one date for everything.

What counts as an intent post?

A public LinkedIn post from someone at the account whose text contains a keyword you defined and whose type carries intent: asking for recommendations, describing a problem, announcing an evaluation or a hire. Reposts, congratulations, and event promotion are dropped. Posters are shown by title, not name, unless you ask.

What does the Skill replace from the prompt gallery?

It packages Find accounts posting about the problem you solve, Check your accounts for leadership changes on the right cadence, Turn company news into outreach, and Get a single-account signal brief before the call into one flow. The prompt pages stay live.

Will it use a lot of Lusha credits?

The Skill calls account_usage first, at zero cost, and states the estimate before pulling signals on more than 10 accounts or intent posts on more than 5. Intent posts are charged per post, which is why the keyword filter runs before the digest is written and the drop count is always shown.

Ready to run this?

One data connection. Works in Claude, ChatGPT, your CRM, or any agent you build.

Start for free Read the Docs