Build a signal library that fires, decays, and combines.
Turns "we should watch for funding" into real signals: how you detect each one, what it is worth, how it decays, which combinations score higher, and the hook that earns the reply. The file that makes your context repo act.
In plain English: “Which signals actually mean now?”
Inside: The detection method, the points, the decay curve, the combination bonuses, the hooks, and a performance-log stub.
Install it in one line, or paste it in.
~/.claude/skills/ and runs automatically when it is relevant.Connect your context. Set it to your motion.
Everything the skill does, in full.
Turns "we should watch for funding" into a real signal library: each signal with how you detect it, what it is worth, how it decays, which combinations score higher, and the message hook that earns the reply. This is the file that makes your context repo act, not just describe. A list of alerts that is never scored or decayed is not a signal library.
- 11. List the moments that precede a deal. The events that mean now, not someday: a new RevOps hire, a Series B in the last 60 days, a usage spike, a competitor renewal coming up. Start with 5 to 8, not 30.
- Give each a detection method. How you actually see it: a Crunchbase alert, a job-board scrape, a product-usage threshold, an intent feed. A signal you cannot detect is a wish.
- Score each in points, against fit. The same signal on a Tier 1 account is worth more than on a Tier 3. Points, not vibes.
- Set a decay curve. A funding round from 150 days ago is not the signal it was at 10 days. Decay the points over time so your active list reflects real urgency, not old history. This one step fixes the list that quietly fills with stale accounts.
- Add combination bonuses. Two signals together tell you more than either alone. A Series B plus a new RevOps hire is not a 65-point account, it is an 80-point account: budget exists and someone is rebuilding. Write the pairs that matter and the bonus each earns.
- Write a message hook per signal. The one line that makes the outreach obviously about them, tied to the signal that fired.
- Leave a performance-log stub. A column for meetings generated per signal, empty for now. Episode 4 fills it, and the library starts to learn which signals actually pay.
- A signal you cannot detect, so it never fires.
- No decay, so the active list is a lie full of six-month-old events.
- Scoring every signal the same regardless of fit, so Tier 3 noise floats to the top.
- Thirty signals nobody maintains instead of eight that work.
[your signal library: 5 to 8 signals with detection, points, decay, combination bonuses, and hooks, plus an empty performance log. This becomes context/signal-library.md]
Example (illustrative):
| Signal | Detection | Points | Decay | Combo | Hook |
|---|---|---|---|---|---|
| New RevOps hire | job-title change | +30 | 90 days | +15 with fresh funding | "Saw you just brought RevOps in-house" |
| Pricing page revisited 3x | product or web analytics | +20 | 14 days | +20 with an active trial | "Noticed the team back on pricing" |
| Series B raised | funding source | +25 | 120 days | +15 with headcount growth | "Congrats on the round" |
Performance log: empty until outcomes come back in Episode 4.
Score the accounts your signals fire on, then wire the freshest one to a motion before it goes stale. The Sales Operator. Receipts only.
Where an operator takes this next.
Defining the signals is the design work. Here's how they start firing on their own.
A signal library that decays and learns is worth more than a longer list of alerts.
Connect Clay or a Crunchbase feed for funding, a job-board scrape for hires, and Amplitude for usage spikes, so each signal fires the moment its detection method sees it instead of you checking manually.
Run a scheduled Claude task nightly that recalculates every account's score against the decay curve, so the active list always reflects real urgency, not last month's snapshot.
Feed booked-meeting outcomes back from Salesforce into the performance log so the library learns which signals actually convert, and the weak ones get retired.
One skill is the on-ramp.
A single skill does one job. Chained into a playbook, or run as a full build, it becomes a system. Here is where this one plugs in.