Terminal
Terminal/Skill
Watches target accounts for funding, M&A, executive hires and other news, and turns each sweep into a dated digest of who to act on.
By Clay
Works with
npx skills add clay-run/clay-skill-creator --skill monitor-buying-signalsThe insight: a buying signal is a dated event you haven't acted on yet — not a fact about a company. The naive build asks "what's in the news about X?" and gets an undated archive: old rounds, recycled press, the same story surfacing every sweep. This skill never asks that question. Every sweep pulls a bounded date window (new since the last look), classifies each event against a menu tied to outbound hook quality, and requires evidence + source + date on every fired signal. And the normal result for most accounts is no signal this window — an honest quiet week beats an invented hook, because a rep who chases one stale or fabricated "signal" stops trusting the whole digest.
Nothing here ships with a value. Each one is the installer's, not the author's: ask for it, never substitute a plausible default, and where an answer does not exist say which step becomes unavailable rather than guessing. Where a default IS defensible it is named below, and using it means saying so in the output.
| Input | What the installer supplies | If it is missing |
|---|---|---|
| The account list | a CSV, table, audience or pasted list with at least a domain column | no default. Dedupe to unique domains first — ten contacts at one account are one watch. Fifty domains is a defensible sweep cap, stated as such |
| The signal menu | which events matter to their motion: funding, exec hires, M&A, expansion, layoffs | funding, M&A and leadership hires is defensible and must be stated. Each chosen signal has to pass one test: would a rep open an email with it? |
| The window | the first-sweep lookback; after that, last sweep to today | 30 days is defensible for the first sweep, and the sweep date goes in every digest |
| Cadence and destination | how often, and where the digest lands | weekly is defensible. The digest goes to a table, a CSV or the conversation — this skill sends nothing anywhere |
If an answer sheet is present beside this skill, load it and ask only for what it does not cover. A partial sheet is normal; a value it is missing gets asked for on its own rather than restarting the interview. Say which values came from the sheet before using them — a sheet applied silently is a wrong field nobody catches. If there is no sheet, say nothing about sheets — the check is a file lookup, not a question, so run the interview as though the feature did not exist rather than reporting an absence. At delivery, offer to save the answers back (identifiers only — never a token or a password), private and never published — and phrase the offer so it explains itself: "want me to save your answers to a file, so the next person on your team doesn't have to answer these again?"
Run clay whoami; echo "exit_code=$?". If it fails or Clay tools are missing, run the
Clay plugin's setup skill, restart if it says to, and re-run this skill. Tell the
user which workspace you're in. Confirm the news arms exist before promising them —
read the live catalog, never your memory of it: clay workflows actions list for the catalog arm and
clay routines get for the managed Company News function — read the declared
costs; they differ by an order of magnitude (see references/news-arms.md).
references/signal-menu.md has the menu with hook-quality ratings.)Two tiers; pick by how standing the watch is, and get explicit approval with the arithmetic stated:
trigger-source, News & fundraising) on an in-app table — the
subscription is free and each landed EVENT costs ~0.5-1 credit, so cost scales with
events, not accounts × sweeps. At roughly weekly cadence on 100+ accounts this is
the cheaper and more timely shape — offer to hand off per
references/news-arms.md §Graduation. Never rebuild the feed by re-scraping —
windowed sweeps are for lists you don't watch continuously.For each unique domain, run the chosen arm with the window pinned ([last sweep → today]; first sweep uses Step 1's lookback). Discipline per call:
Map each event to the signal menu in code — keyword/pattern rules over title +
snippet + date (funding vocabulary, M&A vocabulary, leadership-change titles,
expansion/layoff terms; references/signal-menu.md carries the rule table). An event
matching no menu entry is other — surfaced in the digest tail, never promoted to a
signal. Only genuinely ambiguous events go to an LLM pass, and its verdict must quote
the evidence line it classified from.
Every fired signal emits four fields, none nullable:
signal_present (boolean) · signal_evidence (≤140 chars, from the actual event) ·
signal_source (the event's URL) · signal_date. A signal without a source link or
with a date outside the window is dropped with a note, not shipped.
Digest per account: domain · route (act-now / watch / quiet) · signal type · evidence · source · event date · sweep window, plus a summary: accounts swept, events
found, signals fired by type, quiet count, credits spent (measured where the surface
reports it, declared estimate where it doesn't).
Ask: "Watch my 40 target accounts for funding, M&A, and exec hires — weekly." Dedupe: 40 rows → 34 unique domains. Cost stated: 34 × ~1 credit/sweep (catalog arm), premium arm reserved for the 5 named tier-1 accounts (~33 extra) — user approves the 1-credit arm for all. First sweep, 30-day window:
northwind-systems.example → one event 2026-07-28: "Northwind raises $40M Series B"
→ funding vocabulary match → act-now, evidence "raised $40M Series B led by
Veylmark Capital", source link, opener angle: congrats-on-round.initech-consulting.example → three events, all other (product blog syndication)
→ watch, digest tail.brightloop.example → "Brightloop names new CRO" 2026-08-02 → exec-hire match →
act-now — re-qualify note: confirm the CRO is still in seat and brightloop.example
is the account you target (domain match), then the congrats-on-role angle.
Summary: 34 swept · 5 events · 2 signals fired (1 funding, 1 exec hire) · 31 quiet ·
~34 credits measured from per-call usage. Next sweep's window starts 2026-08-11.