proactive-discovery

name	proactive-discovery
description	Proactive Discovery Engine - Automatically discovers new research opportunities, tracks trends, monitors competitors, and identifies gaps. Transforms the system from reactive to proactive.
version	1.0.0
author	Synthos
license	MIT
allowed-tools	shell (bash, curl, wget), Read (view), Write (write), task_delegation
metadata	{"synthos":{"priority":"P1","atom_type":"cognitive_atom","description":"Proactive discovery - automatically discovers new research opportunities, tracks trends, monitors competitors, and identifies gaps","signature":"domains: list[str] -> discoveries: list[dict], opportunities: list[dict]","related_skills":["knowledge-acquisition","association-discovery","hypothesis-generation","evolution"]}}

Proactive Discovery (主动发现)

Core Positioning

知己知彼，百战不殆。 A super individual does not wait for queries - they discover opportunities. This atom transforms Synthos from a research tool to a research scout.

Goal: Automatically track:

1. New papers in key research domains
2. GitHub trending projects
3. Academic social networks (Twitter/X research threads)
4. Conference deadlines and calls for papers
5. Competitor research output

IO_CONTRACT

• input: domains: list[str] - Research domains to monitor
• input: frequency: str = "daily" - Scan frequency
• output: discoveries: list[dict] - New relevant papers/projects
• output: opportunities: list[dict] - Research opportunities (gaps, trends)
• output: report: str - Human-readable discovery report

Discovery Dimensions

1. Paper Discovery

Scan these sources daily:

• Semantic Scholar API: new papers in VOR, eye movement, biomechanics
• arXiv: cs.AI, cs.LG, physics.med-ph
• PubMed: new vestibular/neurology papers
• Google Scholar alerts (manual setup)

Each new paper:

• Get abstract and metadata
• Compare with existing knowledge base
• Flag if: novel method, novel domain, novel finding
• Queue for ACQ if quality >= threshold

2. Project Discovery

Scan GitHub for:

• Trending repos in AI research agents
• New releases of tools we use (LangGraph, DSPy, PaperQA2)
• Related projects in vestibular/eye tracking
• Open challenges/competitions

Each new project:

• Score quality (1-5)
• If quality >= 4.0, queue for absorption (EVA atom)
• If quality < 3.0, archive as reference only

3. Trend Discovery

Track:

• Citation trends for key papers
• Emerging methods (new ML architectures, new ODE solvers)
• Funding opportunities (NSF, NIH calls)
• Conference deadlines (NeurIPS, ICML, ICLR, IEEE conferences)

4. Gap Discovery

Cross-reference:

• What methods have we used? What methods are emerging?
• What domains have we covered? What domains are adjacent?
• What questions remain unanswered?

Discovery Pipeline

Scan Sources -> Extract -> Score -> Flag -> Queue
     |             |         |       |        |
  API/crawling   metadata  quality  relevant  ACQ/EVA
                & abstract  1-5 score  items   absorption

Output Files

• outputs/discovery/{date}.md - Discovery report
• outputs/discovery/{date}.json - Structured discovery data
• outputs/papers/new-queue/ - Papers queued for processing

Quality Control

• P0 Evidence traceability: Every discovery has source URL/API
• P1 Atomic reproducibility: Each scan independently executable
• P2 Stability sinking: Validated discovery patterns become standard
• P3 Human-machine layering: Human approves which discoveries to pursue

---

善战者，求之于势，不责于人。发现趋势，创造机会。