signal-discovery

name	signal-discovery
description	Guided entry point for defining your ICP and buying signals — runs extract-icp then generate-signals in sequence. Start here before creating signals or building lists.

Signal Discovery

Use this skill as the starting point for signal-based prospecting. It loads your organisation context, then runs extract-icp and generate-signals in sequence to produce a complete, scored signal set ready to activate in Saber.

Saber CLI check

Run saber --help to check if the Saber CLI is installed.

If the CLI is available: proceed with the full workflow below.

If not installed: inform the user that this skill works best with the Saber CLI (available at saber.app), then offer two options:

1. Install the CLI — once installed, restart this skill
2. Continue without the CLI — skip saber org get and ask for org context manually; the ICP extraction and signal generation steps work fully without the CLI

Step 1 — Load organisation context

With CLI: run saber org get

• If the profile contains name, website, and description → use them directly, skip asking.
• If the profile is empty or missing key fields → ask the user to fill in the gaps, then persist:

  saber org update --name "..." --general "..." --products "..." --use-cases "..." --value-prop "..."
  

Without CLI: ask the user:

• What does your company do?
• Who do you sell to?
• What problem do you solve?

Do not ask questions already answered by available context (e.g. CLAUDE.md).

Step 2 — Extract the ICP

Use the extract-icp skill with the company's domain.

This produces a structured ICP covering: target company profile, buying committee, deal economics, pain points, buying triggers, and existing customers.

Keep the ICP output in conversation context for the next step.

Step 3 — Generate signals

Use the generate-signals skill with the ICP from Step 2.

This produces 12–15 research signals across three categories:

• icp_fit — does the company profile match?
• urgency — did a recent trigger event happen?
• buying_signal — is there evidence of the pain?

Each signal includes answer type, weight (1–3), disqualifier flag, and interpretation rules needed for automatic scoring.

Confirm the signal set with the user before handing off. Adjust any signals that don't fit.

Step 4 — Hand off

Keep the agreed ICP and signal definitions in conversation context.

Once signals are approved, tell the user their next steps:

• To build a target list: use build-account-list or build-contact-list
• To activate company signals: use create-company-signals
• To activate contact signals: use create-contact-signals
• To turn the weighted model into native fit + urgency scoring: use configure-scoring (translates categories to dimensions and weights to point values)
• After signals run: use score-accounts to rank accounts by current fit + urgency scores