# Signal Vs Noise Detector

> Separate what actually matters from what is merely loud, recent, vivid, or fashionable. Use when context is cluttered, too many inputs compete, or the team needs to know which signals deserve attention and which should be ignored.

- Skill: `alubiama/signal-vs-noise-detector` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add alubiama/signal-vs-noise-detector`
- Raw SKILL.md: https://api.skillmd.com/api/skills/alubiama/signal-vs-noise-detector/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Alubiama (https://skillmd.com/u/alubiama)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/alubiama/signal-vs-noise-detector

---


# Signal vs Noise Detector

## Goal

Reduce clutter by identifying the few signals that should drive the next decision.

This skill does not summarize everything. It filters inputs so focus goes to the highest-value information.
It does not verify whether a signal is true in depth and it does not decide the final overall verdict by itself.

## When To Use

- too many inputs compete
- recent or loud information is distorting judgment
- research or brainstorming produced clutter
- the next decision needs a clear signal set

## Scope Boundaries

In scope:

- classify inputs by decision relevance
- separate signal from distraction
- name what to ignore for now

Out of scope by default:

- full synthesis documents
- proving the final truth of each input
- replacing `evidence-gap-detector`, `quality-gate`, or `priority-engine`

## Escalation Conditions

Pause and keep the filter provisional when:

- the decision context itself is still unclear
- input quality is too weak to separate relevance from noise
- the real issue is evidence quality, not information clutter
- ranking the signals requires a priority call that has not been made yet

If filtering stays unstable, route to `intent-clarifier`, `evidence-gap-detector`, or `priority-engine` instead of forcing a neat signal set.

## Workflow

1. State the decision to support.
2. List the current inputs or observations.
3. Score lightly for:
   - relevance
   - reliability
   - uniqueness
   - action impact
4. Keep the strongest signals.
5. Mark the rest as noise, secondary, or deferred.
6. Limit the primary focus set to the smallest useful number.
7. Recommend the focus set.

## Output Contract

Always return:

1. `decision_context`
2. `input_set`
3. `signal_set`
4. `noise_set`
5. `secondary_signals`
6. `why_signal_wins`
7. `what_to_ignore_for_now`
8. `recommended_followup`
9. `next_step`

## Guardrails

- Do not confuse loudness with relevance.
- Do not keep low-quality inputs just because they are recent.
- Prefer a small signal set.
- Keep secondary-but-useful inputs separate from true noise when that distinction matters.
- Default to at most 3 primary signals unless the decision clearly requires more.

## Anti-Patterns

- everything is signal
- recency bias
- fashionable noise