| name | omega-research |
| description | Structured research pipeline for Omega — takes an idea or hypothesis, searches for evidence, compares to existing Omega capabilities, and delivers a verdict with implementation recommendations. Use when asked to /research a topic, evaluate a new signal source, assess an external tool/library, or investigate trading strategies. |
| tags | ["research","omega","victoria","strategy","analysis"] |
Omega Research Skill
Run structured research on any idea, signal, tool, or strategy and produce an actionable verdict with Omega integration notes.
Trigger phrases
/research <topic>
- "research X"
- "look into X for Omega"
- "evaluate X for Victoria"
- "should we add X to the pipeline?"
The Pipeline
IDEA → DECOMPOSE → SEARCH → COMPARE-TO-OMEGA → VERDICT → LOG
Phase 1: Decompose
Break the research topic into 3–5 sub-questions:
- What is it and how does it work?
- What is the claimed edge / advantage?
- What does the literature / community say about real-world performance?
- Is there existing Omega/Victoria overlap or conflict?
- What would implementation cost (time, API cost, infra)?
Phase 2: Search
Use WebSearch on each sub-question. Prefer:
- arXiv, SSRN for academic signals
- GitHub for implementations
- Twitter/X via SN13 feed for practitioner alpha
- Polymarket blog / forecast community for prediction market angles
Collect 3–5 authoritative sources per sub-question.
Phase 3: Compare to Omega
Check the following files for overlap:
omega-strategic-backlog.md — is this already tracked?
omega/nodes/ — is there an existing node that covers this?
docs/DATA_SOURCES.md — is the data source already integrated?
docs/specs/ — any relevant spec?
Explicitly state: NEW, OVERLAP (with item ID), or SUPERSEDES (item ID).
Phase 4: Verdict
Produce a structured verdict block:
## Research Verdict: <topic>
Date: YYYY-MM-DD
### TL;DR
One sentence conclusion.
### Evidence Quality
[HIGH / MEDIUM / LOW] — explain why
### Omega Overlap
[NEW / OVERLAP: <ID> / SUPERSEDES: <ID>]
### Recommended Action
- [ ] Add to backlog as <ID>: <description>
- [ ] Implement immediately (reason)
- [ ] Discard (reason)
- [ ] Monitor (next review date)
### Implementation Notes
- Relevant node: <NodeName>
- Data source: <API/SDK>
- Estimated cost: <$/mo>
- Estimated effort: <S/M/L>
### Sources
1. <url> — <one-line summary>
2. ...
Phase 5: Log
Append the verdict to docs/ideas/feed.jsonl in this format:
{"timestamp": "ISO8601", "type": "research_verdict", "topic": "...", "action": "add_backlog|implement|discard|monitor", "backlog_id": "...", "summary": "...", "sources": [...]}
If high-priority (implement immediately), also print a [PRIORITY] banner.
Rules
- Always complete all 5 phases before responding
- Never skip Phase 3 — Omega overlap check prevents duplicate work
- Cost estimates are required for any external API
- Log entry is mandatory — this feeds the scheduled research task