Last used: 2026-03-24
Memory references: 1
Status: Active
Fact-Checker: Verify Markdown Claims Against Source Data
Given a markdown draft file, this skill extracts every verifiable claim
(numbers, dates, model names, scores, causal statements) and cross-references
them against available source data to produce a verification report.
Usage
CODEBLOCK0
What It Checks
Claim types extracted
- - Numeric claims — integers and floats with surrounding context
- Model references —
model/task (phi4/classify) and model:tag (phi4:latest) - Dates —
YYYY-MM-DD format - Score values — decimal scores like
0.923, INLINECODE4 - Percentages —
42%, INLINECODE6
Source data consulted (in priority order)
- 1.
projects/hybrid-control-plane/FINDINGS.md — primary source of truth - Control Plane
/status API at http://localhost:8765/status — live scored run data - INLINECODE10 — raw scored run files on disk
- INLINECODE11 — daily logs with timestamps and decisions
- INLINECODE12 in
projects/hybrid-control-plane/ — commit hashes, dates, authorship - INLINECODE14 — sprint history
Output Format
Each claim produces one line:
CODEBLOCK1
Followed by a summary count of confirmed / unverifiable / contradicted claims.
When To Use This Skill
When asked to "fact-check" or "verify" a draft blog post, report, or
documentation file — run this skill and present the report to the user.
If any claims are ❌ CONTRADICTED, flag them prominently and suggest corrections.
Instructions for Agent
- 1. Run the script with the path to the draft file.
- Parse the output report.
- Summarise key findings — especially any ❌ CONTRADICTED claims.
- Suggest specific corrections with the correct values from the evidence.
- If the
/status API is unavailable, note it and rely on FINDINGS.md + score files.
技能名称: fact-checker
详细描述:
最后使用时间: 2026-03-24
记忆引用: 1
状态: 活跃
事实核查器:根据源数据验证Markdown声明
给定一个Markdown草稿文件,此技能会提取每个可验证的声明(数字、日期、模型名称、分数、因果陈述),并将其与可用的源数据进行交叉引用,以生成一份验证报告。
使用方法
bash
python3 skills/fact-checker/scripts/fact_check.py
python3 skills/fact-checker/scripts/fact_check.py --output report.md
检查内容
提取的声明类型
- - 数字声明 — 整数和浮点数及其上下文
- 模型引用 — model/task(phi4/classify)和 model:tag(phi4:latest)
- 日期 — YYYY-MM-DD 格式
- 分数值 — 如 0.923、1.000 这样的小数分数
- 百分比 — 42%、95.3%
参考的源数据(按优先级排序)
- 1. projects/hybrid-control-plane/FINDINGS.md — 主要事实来源
- http://localhost:8765/status 上的控制平面 /status API — 实时评分运行数据
- projects/hybrid-control-plane/data/scores/.json — 磁盘上的原始评分运行文件
- memory/.md — 带有时间戳和决策的每日日志
- projects/hybrid-control-plane/ 中的 git log — 提交哈希、日期、作者
- projects/hybrid-control-plane/CHANGELOG.md — 冲刺历史记录
输出格式
每个声明生成一行:
✅ 已确认: phi4/classify scored 1.000 → /status API:phi4latestclassify mean=1.000 n=23
⚠️ 无法验证: this took about a day → 日志中未找到时间戳关联
❌ 矛盾: 909 runs → /status API 显示总共 958 次运行(数字过时?)
随后是已确认/无法验证/矛盾声明的汇总计数。
何时使用此技能
当被要求事实核查或验证一篇草稿博客文章、报告或文档文件时——运行此技能并向用户展示报告。如果有任何声明是 ❌ 矛盾的,请突出标记并建议更正。
代理指令
- 1. 使用草稿文件的路径运行脚本。
- 解析输出报告。
- 总结关键发现——尤其是任何 ❌ 矛盾的声明。
- 根据证据中的正确值建议具体的更正。
- 如果 /status API 不可用,请注明并依赖 FINDINGS.md 和评分文件。