fact-check-x-knowledge-compare
在本地把多个 AI 原始回答拆成原子知识点,区分直接答案与补充参考,对齐各平台主张、差异、所附依据和来源忠实性,并生成可视化中间报告。不调用可信搜索或任何外部模型 API,语义分析由当前运行载体完成。
Skill by ASI2030
Chalk Labs
Need help deploying fact-check-x-knowledge-compare?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/asi2030-fact-check-x-knowledge-compare
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: backend
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 26, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Python package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/fact-check-x-knowledge-compare
What's inside
- Skill name
fact-check-x-knowledge-compare- Description
- 在本地把多个 AI 原始回答拆成原子知识点,区分直接答案与补充参考,对齐各平台主张、差异、所附依据和来源忠实性,并生成可视化中间报告。不调用可信搜索或任何外部模型 API,语义分析由当前运行载体完成。
- License
- Apache-2.0
- Instructions
- 60 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (15)
- SKILL.md
- openai.yaml · agents/
- acceptance-criteria.md · references/
- contracts.md · references/
- common.py · scripts/
- knowledge_compare.py · scripts/
- policy_match.py · scripts/
- render_comparison.py · scripts/
- citation_presence_test.py · tests/
- deepseek_reference_faithfulness_test.py · tests/
- comparison-analysis.json · tests/fixtures/
- results.json · tests/fixtures/
- model_judgment_test.py · tests/
- requirements_scope_test.py · tests/
- smoke_test.py · tests/
SKILL.md (first lines)
# 知识点结构化对比
本技能是独立、可选的本地中间层。它回答“各家说了什么、哪里相同或不同、各自有没有依据、是否忠实于自己的依据”,不回答外部世界中谁最终正确。
当前承载技能的智能体(Codex、Claude Code 等)直接完成知识点拆解和语义判断。脚本只生成任务包、校验证据约束、标准化结构并渲染报告,不调用模型 API。
## 第一步:生成分析任务包
```bash
python3 scripts/knowledge_compare.py \
--input <run>/results.json \
--task-output <run>/comparison-task.json
```
阅读 `comparison-task.json` 后,由当前智能体写出 `comparison-analysis.json`。必须遵守:
- 先读紧凑任务包。`capturedText` 只是预览;预览不够时按 `sourceResultsPath` 回读该来源全文,不限制为固定三次工具调用。
- 来源只有平台片段时,由模型判断是否需要补全同一原来源;可用采集器 `hydrate-sources` 命令。保留原结果,不换来源、不重提问题;补全后从新采集版本重新比较。
- 合并所有成功平台中可独立核验的事实,每个知识点只表达一个事实变量。
- 原子性同时约束知识点和各平台 `claim`;原句包含多个独立义务、条件、对象、数值或后果时必须拆点,每个 `claim` 只保留当前事实,`answerExcerpt` 可以复用原句。
- 金额换算、比较符号、简称、算术推导和基准/优惠关系由模型结合原文理解,在理由中说明。是否拆点取决于业务含义,不用固定数值、关键词或字符串相似度决定真假。
- 仅个别平台提出的实质新增事实也要单独成点,其他平台标为未覆盖;不得把新增事实并入宽泛知识点后借用深知晓锚点免查。
- 同一事实的不同数值或条件放在同一知识点下,不拆成两点。
- `direct` 采用最小充分原则:缺失后就没有直接回答用户明确所问内容;`reference` 是背景、提醒、延伸或相邻业务信息。问全国时省市案例、问城市时区县细项均属于补充参考,除非用户明确询问地区差异;只问条件/门槛/比例时,奖励、流程、电话等不进入直接答案。
- 每个平台逐点写明主张、所附依据和来源忠实性。
- 优先保留实际脚标对应关系。引用其他回答级材料时如实标明,不伪装为局部脚标,不改写原来源身份。
- 同一平台的全局来源清单可作为判断材料;是否足以支撑当前主张由模型说明,不因缺少局部脚标自动判巧合。
- 比较阶段使用当前捕获的原答案和来源;补全也只针对同一原来源。独立检索新证据属于后续事实核验阶段,不能倒填为平台自带引用。
- 若确认深知晓该知识点使用可信搜索且所附官方证据支持同一个原子事实,可声明 `trustedAnchor`;锚点只能覆盖该事实变量,超出部分必须另起无锚点知识点,由下游可信搜索核验。
- 同一知识点若普通深知晓和深度溯源各自附有支持其主张的官方原文,必须全部保留到 `trustedAnchor.claimEvidenceMap`;后续权威核验直接复用,不得因追加搜索未召回同一材料而降为证据不足。
- 顶层必须生成 `synthesisDraft`:`status` 固定为 `unverified`,正文基于全部相关知识点合并并保留冲突、条件和缺口,`basisKnowledgePointIds` 只能引用当前知识点。它是独立可交付的“综合草案(未核验)”,不能冒充权威最终答案。
## 第二步:验收并渲染
```bash
python3 scripts/knowledge_compare.py \
--input <run>/results.json \
--analysis <run>/comparison-analysis.json \
--output <run>/comparison.json
python3 scripts/render_comparison.py \
--results <run>/results.json \
--comparison <run>/comparison.json \
--output <run>/comparison.html
```
`comparison.html` 按实际平台数展示全部原始答案与来源、知识点对比概览和
“综合草案(未核验)”;下方展示逐知识点完整对照。
## 交付门禁
```bash
python3 tests/smoke_test.py
python3 "${CODEX_HOME:-$HOME/.codex}/skills/.system/skill-creator/scripts/quick_validate.py" .
```
数据形状见 [数据契约](references/contracts.md),分类和引用规则见 [验收标准](references/acceptance-criteria.md)。From ASI2030/Fact-Check-X, Apache-2.0.
More from ASI2030/Fact-Check-X
- 0.611score
fact-check-x-authoritative-verify
ASI2030
对单个知识点调用可信搜索取得官方证据,由当前运行载体裁决各平台主张,并分别生成权威核验后的最终答案与各方答案测评报告。支持多知识点独立并发、深知晓官方材料锚点免重复检索,以及直接准确、间接准确、结果巧合、严重误导、凭空编造等分类。可用于 Fact-Check-X 完整流程,也可作为独立的权威证据核验能力使用。
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
fact-check-x-complete
ASI2030
面向真实业务决策的多平台事实核验 Skill。正式支持深知晓、深知晓(深度溯源)、豆包、腾讯元宝、DeepSeek 和通义千问,按用户输入动态选择 N≥1 个平台。完整采集原回答、引用、截图与页面存证,结构化比较关键事实,按需使用可信搜索逐点核验权威证据,并评估各平台表现。权威答案是核验产物之一,原始证据、冲突与证据…
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
llm-answer-reference-compare
ASI2030
用 Playwright 从深知晓、深知晓(深度溯源)、豆包、腾讯元宝、DeepSeek 和通义千问网页端无损采集完整回答、原始引用、引用标记和现场存证。用于 Fact-Check-X 的原始答案采集阶段,或任何需要比较多个 AI 网页回答与参考来源的任务。
SkillAFreshness: A★ 4Claude Code +9 - 0.608score
fact-check-x-unified
ASI2030
Fact-Check-X 流程编排能力,依次组织各方答案汇总、各方答案聚合(未核验)、权威核验后的最终答案和各方答案测评,生成可打开、可审计、可迁移的阶段产物与完整报告包。
SkillAFreshness: A★ 4Claude Code +9
Signals
- Stars
- 4
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 19, 2026 (11 days ago)
- Last release
- Oct 19, 2026 (11 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/ASI2030/Fact-Check-X
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from ASI2030
- 0.611score
fact-check-x-authoritative-verify
ASI2030
对单个知识点调用可信搜索取得官方证据,由当前运行载体裁决各平台主张,并分别生成权威核验后的最终答案与各方答案测评报告。支持多知识点独立并发、深知晓官方材料锚点免重复检索,以及直接准确、间接准确、结果巧合、严重误导、凭空编造等分类。可用于 Fact-Check-X 完整流程,也可作为独立的权威证据核验能力使用。
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
fact-check-x-complete
ASI2030
面向真实业务决策的多平台事实核验 Skill。正式支持深知晓、深知晓(深度溯源)、豆包、腾讯元宝、DeepSeek 和通义千问,按用户输入动态选择 N≥1 个平台。完整采集原回答、引用、截图与页面存证,结构化比较关键事实,按需使用可信搜索逐点核验权威证据,并评估各平台表现。权威答案是核验产物之一,原始证据、冲突与证据…
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
llm-answer-reference-compare
ASI2030
用 Playwright 从深知晓、深知晓(深度溯源)、豆包、腾讯元宝、DeepSeek 和通义千问网页端无损采集完整回答、原始引用、引用标记和现场存证。用于 Fact-Check-X 的原始答案采集阶段,或任何需要比较多个 AI 网页回答与参考来源的任务。
SkillAFreshness: A★ 4Claude Code +9 - 0.608score
fact-check-x-unified
ASI2030
Fact-Check-X 流程编排能力,依次组织各方答案汇总、各方答案聚合(未核验)、权威核验后的最终答案和各方答案测评,生成可打开、可审计、可迁移的阶段产物与完整报告包。
SkillAFreshness: A★ 4Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get fact-check-x-knowledge-compare running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
fact-check-x-knowledge-compare skill. Agent Mart. https://agentmart.dev/skills/asi2030-fact-check-x-knowledge-compare (last updated 2026-10-24)