AGENT SKILL

research

Multi-source research & knowledge synthesis with cross-validation, provenance graph, confidence score

License 许可
MIT
最近更新
2026-11-02
Stars
★ 1
Forks
0

01 前言 Front Matter

base
skills/fetching
name
research
version
1.0.0
category
Research/Knowledge
provides
[ "research.run" ]
security
{ "inherit": "fetching", "ssrf_https_only": true }
description
Multi-source research & knowledge synthesis with cross-validation, provenance graph, confidence score

02 资源文件 4

03 SKILL.md 正文

Research — Deep Multi-Source Knowledge Synthesis

Synthesize knowledge across many sources: fetch → dedupe → cross-validate → confidence score + provenance graph.

Usage

from skills.research import research
res = await research("FastAPI best practices",
                     sources=["https://api.example.com/a", "https://api.example.com/b"])
print(res["summary"])      # Synthesized 2 source(s) on ... — NN% confidence
print(res["confidence"])   # 0.0–1.0
print(res["provenance"])   # graph + fetched_at + checksum

Return shape

{
  "summary": "str",
  "confidence": 0.87,
  "contradictions": [],
  "sources_used": ["https://..."],
  "sources_skipped": [],
  "provenance": { "graph": [...], "fetched_at": "ISO", "cache": "MISS", "checksum": "sha256:..." }
}

Security

All outbound traffic routes through skills.fetching, inheriting SSRF guard, HTTPS-only, retry, and cache. localhost/private/metadata hosts are skipped.

04 同仓库更多 Skills