Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

signal-first-research

Ask DeepWiki GitMCP

An Agent Skill for designing research intake filters that admit only information likely to change your next action. Most "stay current" workflows fail either by volume (a guilt queue that outruns reading time) or by false-abstinence (reading nothing because sorting feels too expensive). A signal-first filter refuses both: search widely, intake narrowly.

Install

Claude Code

# Copy into your global skills directory
cp -r skills/signal-first-research ~/.claude/skills/signal-first-research

SkillsMP

/skills add shimo4228/signal-first-research

How It Works

  1. Define the signal before the search — write down what information would actually change your next action
  2. Search without source exclusion — no source is filtered upfront; breadth stays intact
  3. Filter at intake, not at read-time — each item must answer "would this change what I do next?" before it is admitted
  4. Diagnose the filter periodically — a checklist distinguishes a healthy filter from one that has quietly broken (lazy passthrough, topic creep, guilt backlog)

When It Triggers

  • You are about to build a recurring research workflow — daily digest, news stream, topic monitor, literature feed
  • The default question "what should I read?" keeps producing too much
  • An existing digest has become a backlog you skim with guilt instead of acting on

Failure Modes It Prevents

Failure Shape Signal-first answer
Volume bias Aggregate everything, trust yourself to skim Filter lives in the intake question, not in your head
False-abstinence Read nothing; sorting feels too expensive Breadth is kept; only intake is narrowed
Lazy filter Filter passes everything "interesting" "Interesting" is not a signal; action-changing is

Syncing from the harness

The canonical copy of this skill lives in the author's live Claude Code harness. This repository is a one-way publication mirror:

scripts/sync-from-local.sh --dry-run   # report differences only
scripts/sync-from-local.sh             # apply to working tree (never commits)

About this skill

This skill is a design-pattern skill from the Agent Knowledge Cycle (AKC) research line — a Zenodo-citable six-phase bidirectional growth loop (DOI 10.5281/zenodo.19200726) for sustaining intent alignment between an AI agent and its operator over time. It is the "how" counterpart to AKC ADR-0010 Human Cognitive Resource as Central Constraint. AKC is one of three research lines by @shimo4228, alongside Contemplative Agent (DOI 10.5281/zenodo.19212118) — autonomous agents grounded in four contemplative axioms — and Agent Attribution Practice (AAP) (DOI 10.5281/zenodo.19652013) — harness-neutral ADRs on accountability distribution.

License

MIT

About

Agent Skill: design a research intake filter that admits only information likely to change your next action. Search widely, intake narrowly.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages