| name | research-intelligence |
| description | Use when gathering football, sports, fantasy, or prediction-market intelligence from multiple sources, especially when the work requires source comparison, freshness checks, provenance tracking, and extraction of structured evidence for downstream agents in any runtime. |
Research Intelligence
Use this skill when the task is to gather, compare, and structure evidence rather than jump straight to a recommendation.
Goals
- Find the highest-signal sources first.
- Preserve source provenance on every important claim.
- Separate facts, rumors, opinions, and model inferences.
- Produce structured evidence that later agents can reuse.
Workflow
- Define the decision target.
- Identify the minimum entities needed: competition, fixture, teams, players, market, and deadline.
- Gather fresh sources, prioritizing official or primary sources.
- Extract discrete evidence items instead of long summaries.
- Score each evidence item for freshness, reliability, and relevance.
- Highlight conflicts and unresolved gaps.
Output format
Prefer returning evidence in a consistent shape:
{
"claim": "",
"entity": "",
"source": "",
"timestamp": "",
"reliability": "high|medium|low",
"freshness": "fresh|aging|stale",
"type": "fact|report|opinion|inference",
"impact": "",
"notes": ""
}
Important rules
- Never flatten conflicting reports into fake certainty.
- Explicitly label rumors and unconfirmed lineup news.
- Prefer concise evidence bullets over narrative prose.
- If a source is derivative, try to find the original upstream source.
- When browsing current information, use concrete dates and times.
When to escalate
Escalate to the orchestrator when:
- sources disagree on a high-impact issue,
- freshness is poor near deadline,
- multiple entities are unresolved,
- a recommendation would be premature.
