research

Multi-Agent Research

Wave-based knowledge gathering with deferred synthesis. Mined from 300+ real research dispatches: the pattern that consistently produces actionable intelligence.

Core insight: Research breadth-first, synthesize after. Conclusions drawn from the first three results miss nuance the fourth wave would have surfaced. Deploying agents in waves and accumulating findings before synthesizing produces sharper recommendations.

How to read this skill: calibrate to the question, not the framework. Most research is Quick Mode or one wave plus targeted follow-ups; only deep dives need the full pattern. The Phase 1 budget table sizes it.

The Shape

digraph research {
    rankdir=TB;
    node [shape=box];

    "1. PRIME" [style=filled, fillcolor="#e8e8ff"];
    "2. WAVE 1: Broad Sweep" [style=filled, fillcolor="#ffe8e8"];
    "3. GAP ANALYSIS" [style=filled, fillcolor="#fff8e0"];
    "4. WAVE 2+: Targeted" [style=filled, fillcolor="#ffe8e8"];
    "5. SYNTHESIZE" [style=filled, fillcolor="#e8ffe8"];
    "6. DECIDE & RECORD" [style=filled, fillcolor="#e8e8ff"];

    "1. PRIME" -> "2. WAVE 1: Broad Sweep";
    "2. WAVE 1: Broad Sweep" -> "3. GAP ANALYSIS";
    "3. GAP ANALYSIS" -> "4. WAVE 2+: Targeted";
    "4. WAVE 2+: Targeted" -> "3. GAP ANALYSIS" [label="still gaps", style=dashed];
    "3. GAP ANALYSIS" -> "5. SYNTHESIZE" [label="coverage sufficient"];
    "5. SYNTHESIZE" -> "6. DECIDE & RECORD";
}

---

Phase 1: PRIME

Lean on existing knowledge before spawning agents. Re-running research that already lives in Sibyl burns tokens and produces duplicate entries.

Common moves

• Search Sibyl first: sibyl search "<research topic>", sibyl search "<related technology>", sibyl search "<prior decision in this area>". Surface what's already known before generating new findings.
• Check for staleness. Fast-moving topics (frameworks, models, cloud services) usually warrant re-research even when Sibyl has recent entries; treat the existing knowledge as a baseline. Stable topics with recent entries often don't need a fresh pass at all. One class is never exempt: version, "latest", and SOTA facts expire no matter how recent the entry feels. Recalled memory routes the investigation, live state decides. Re-verify those against the primary source before they drive a dispatch or a recommendation. Same rot law for prior research docs: anything older than the reality it describes gets a per-claim liveness check against live sources and current code before it shapes a decision.
• Premise-check the target. Confirm the data, repo, or question actually exists (and disambiguate which one) before any agent launches. A wave pointed at a wrong or empty target manufactures findings.
• Sharpen the research question. "Research databases" is too vague to dispatch on. "Compare PostgreSQL vs CockroachDB for multi-region write-heavy workloads with <10ms p99 latency" gives agents enough scope to do useful work.
• Calibrate the research budget to the decision the research is feeding: