extract-signal

Extracts and vets signal from noisy sources using composable sub-agent techniques.

14|1|Updated Jun 14, 2026
One-click install
npx skills add https://github.com/allemaar/open-skills --skill extract-signal-allemaar
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: extract-signal
Source: https://github.com/allemaar/open-skills/tree/main/skills/extract-signal
Command: npx skills add https://github.com/allemaar/open-skills --skill extract-signal-allemaar

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Gathering reliable answers from noisy sources (large codebases, diffs, many documents) pollutes the working context and often produces unvetted claims. This Skill relocates the noisy gathering work into a disposable sub-agent window and returns only a vetted, provenance-tagged signal in the form the requester asked for. ## Core Features & Use Cases - Elastic technique composition: Infers the signal, output form, stakes, and destination from the call, then composes gather (/prime-sweep, /prime-fetch, /prime-expand), diverge (/insight-explore), and vet (/insight-critique, /insight-assess, cold second opinion) techniques to fit the task. - Context hygiene by relocation: Defaults to spawning a freshly-primed sub-agent so only the signal reaches the caller's window; works inline when it is itself the running agent, with a depth-1 guardrail. - Provenance and sensitivity contracts: Every presented signal carries source:line provenance, confidence, residual gaps, and techniques-run, and is sensitivity-matched to its destination so internal detail never leaks into public-bound output. - Use Case: Ask "is this package ready to ship publicly?" and receive a bill-of-health report with per-area findings, source references, confidence levels, and stated gaps, produced by a vetted sub-agent pipeline. ## Quick Start Ask the agent to extract the signal on whether your module is ready to ship and present it as a bill-of-health report.

Frequently Asked Questions about extract-signal

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract a vetted signal from a noisy codebase or document set?

Invoke the skill with a natural-language request like "extract the signal on X and present it as Y". It infers the question, output form, and stakes, then composes gathering and vetting techniques and returns a provenance-tagged result.

How does sub-agent extraction keep my working context clean?

By default the skill spawns a freshly-primed sub-agent that performs the noisy gathering and vetting in its own disposable window. Only the final vetted signal returns to the caller, so the main context is never flooded with raw source material.

What output forms can signal extraction produce?

Supported forms include verdicts, ranked options, risk lists, maps or inventories, counts or metrics, and bills of health, plus any custom form. The form is inferred from phrasing such as "is it ready?" or "what are the options?".

Does the skill prevent internal details from leaking into public output?

Yes. The signal is tagged with a sensitivity level and matched against its destination. If output is bound for a lower-trust or public destination, sensitive parts are withheld and the caller is told what was withheld and why.

What happens when vetting techniques disagree or gathering finds nothing?

Disagreements between assessment and the cold second opinion are surfaced with both positions rather than averaged into false consensus. An empty gather is reported as "no signal found" with what was searched; the skill never fabricates signal.