Research
zhongkaifu/TensorSharp
A skill your agent uses for web searches and current information lookups, finding sources, fact-checking, researching questions, comparing sources, or summarising web pages.
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、冲突与证据边界全程可追溯。基础采集与对比无需 API…
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install aiskillstore/marketplace fact-check-x-complete --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .claude/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .claude/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-completeType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install aiskillstore/marketplace fact-check-x-complete --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .agents/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .agents/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install aiskillstore/marketplace fact-check-x-complete --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .cursor/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .cursor/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/aiskillstore/marketplace.git --path skills/asi2030/fact-check-x-complete--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install aiskillstore/marketplace fact-check-x-complete --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .gemini/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .gemini/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install aiskillstore/marketplace fact-check-x-completeInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .github/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .github/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install aiskillstore/marketplace fact-check-x-complete --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/aiskillstore/marketplace.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/asi2030/fact-check-x-complete .opencode/skills/fact-check-x-complete && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "fact-check-x-complete" agent skill from https://github.com/aiskillstore/marketplace/tree/main/skills/asi2030/fact-check-x-complete into .opencode/skills/fact-check-x-complete/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "fact-check-x-complete", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
fact-check-x-complete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、冲突与证据边界全程可追溯。基础采集与对比无需 API…
Fact Check X Complete is an agent skill from aiskillstore/marketplace. 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、冲突与证据边界全程可追溯。基础采集与对比无需 API Key;可信搜索是权威核验阶段的可选增强。语义工作由 Codex、Claude Code、WorkBuddy 等当前运行载体完成,不调用外部大模型 API。
Its SKILL.md is about 3.8k tokens, which your agent loads only when the skill is triggered. The skill folder holds 124 other files, including scripts and reference files (for example `PRIVACY.md`, `SECURITY.md` and `THIRD_PARTY_NOTICES.md`).
It sits in Research & Science, covering Fact-checking and source verification. It works with DeepSeek. The repository describes itself as: Security-audited skills for Claude, Codex & Claude Code. One-click install, quality verified. The licence is Apache-2.0.
10 steps, taken from the first numbered list in SKILL.md.
Read from SKILL.md and the folder at commit 44923f3. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Ships 1 file in scripts/ (Python, from the files we listed), which the agent can run.
Shell commands in SKILL.md call:
python3nodenpmFrom the folder's file list and the shell code blocks in SKILL.md.
Links to these hosts (documentation or services it may open):
raw.githubusercontent.comFrom URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Fact Check X Complete loads about 3.8k tokens when it runs, and up to ~10k if it reads all its reference files. Until then it costs about 70 tokens; SKILL.md has 564 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.
The full file from aiskillstore/marketplace at commit 44923f3, republished under its Apache-2.0 licence (© aiskillstore). 564 words, ~3,781 tokens.
.claude/skills/fact-check-x-complete/SKILL.md (or your agent's skills folder). This skill also uses 118 other files; get the full folder from GitHub.Image: Fact-Check-X 多平台事实核验:完整采集、知识点对比、权威核验与答案生成、平台表现评估
把同一个问题交给一个或多个 AI 平台,完整保留每家的回答和引用,再把关键事实逐点对齐、核验并评估各平台表现。第三步会基于权威证据生成最终答案,但不会用“答案生成”代替完整事实核验:证据冲突会被保留,证据不足的内容不会写入确定结论。用户只需说出问题和要比较的平台,不需要学习平台 ID、内部流程编号或报告术语。
当前内置以下网页端平台,按用户本次输入动态选择任意 N≥1 个:
| 平台 | 能力 |
|---|---|
| 深知晓 | 标准问答、引用与官方来源采集 |
| 深知晓(深度溯源) | 等待普通回答完成后启动深度研究,并作为独立结果采集 |
| 豆包 | 回答、引用与页面存证 |
| DeepSeek | 回答、引用与页面存证 |
| 通义千问 | 回答、引用与页面存证 |
| 腾讯元宝 | 回答、引用与页面存证 |
平台由用户选择,不存在固定“五平台模式”。N=1 可完成单平台事实核验;N≥2 会额外比较各平台的共识、冲突和来源差异。网页结构变化时,技能会明确报告失败并停在当前阶段,不会用部分结果冒充完整报告。 |
login 验证可见浏览器,但没有真实成功输出前不得声称浏览器已启动。login、run、prepare-comparison、complete-comparison、prepare-authority、search-authority、finalize-authority 和 deliver。执行工具因等待用户操作而返回可轮询的运行会话 ID 时,应保留并轮询该会话;禁止使用 shell 后台任务,禁止给任何流水线命令添加 | tail、| tee、|| true 或其他会掩盖退出码的包装。login 是可见浏览器命令,必须严格按本技能给出的参数执行,不得添加 --headed 或其他未列出的参数。精确命令尚未失败时,不得先检查 CLI 帮助、源码或浏览器环境;精确命令失败后只按错误与 capture-recovery.json 恢复。login 成功、可提问页面已建立或 results.json 全平台成功证据前,禁止告诉用户“浏览器已启动”“采集已在后台运行”或“将自动进入后续流程”。login 或 run 运行期间出现登录、短信验证码、人机验证或 CAPTCHA 时,保持命令和当前 Playwright 页面运行,立即提示用户本人处理;检测到处理完成后自动继续,不得关闭页面、重输问题或提前结束任务。login 或 run 非零退出、浏览器意外关闭、人工处理超时或采集失败时,立即读取 capture-recovery.json。运行载体有 Computer Use 时调用它恢复;没有时明确说明“当前载体无法调用 Computer Use,原始答案采集已停止”,停在原始答案采集阶段。capture-recovery.json.question。capture-gate.json 未证明所有指定平台均成功前,禁止进入知识点对比和后续流程。不得用已有材料、搜索结果、空回答或部分成功结果替代失败平台。本技能是对外唯一入口,包内自带三个独立业务模块:
llm-answer-reference-compare:多端回答、原始引用与现场存证无损采集。fact-check-x-knowledge-compare:原子知识点拆解、主张对齐、来源忠实性判断。fact-check-x-authoritative-verify:逐知识点可信搜索、证据裁决、独立权威核验报告与最终平台表现报告。脚本只负责采集、校验、编排和渲染。知识点拆解、证据理解与裁决由当前运行载体完成,禁止调用任何外部模型 API。
所有面向用户的提示、阶段检查点、错误说明和最终答复必须使用中文。不得因为运行载体的默认语言改用英文。
先定位包内模块并自动准备采集运行时:
python3 scripts/fact_check_x.py locate
python3 scripts/fact_check_x.py prepare-runtimeprepare-runtime 会先自检;依赖完整时直接跳过,首次缺失时根据包内锁文件自动执行
npm ci --omit=dev,并在实际调用采集 CLI 验证通过后才返回。不得借用其他技能目录或
旧安装残留的 node_modules,也不要求用户手工进入目录安装依赖。
正常在线采集直接使用系统 Chromium 浏览器,不要求下载 Chrome for Testing。 只有无头/CI 回归确有需要时才单独安装 Playwright 测试浏览器。
macOS 可见登录与交互采集按以下顺序自动选择系统浏览器:用户通过
FACT_CHECK_X_BROWSER_EXECUTABLE 指定的 Chromium 浏览器、Google Chrome、
Microsoft Edge、Brave、Chromium。Playwright 自带的 Chrome for Testing
默认只用于无头/CI 回归。系统没有上述浏览器时,使用 Computer Use 或提示用户
安装任一受支持浏览器;不得反复启动 Chrome for Testing。
首次采集某个平台时,必须先在对话中告诉用户“将打开浏览器,请完成登录;检测到可提问界面后自动保存会话”。不得把提示只藏在命令输出里。然后逐个平台打开可见浏览器并等待登录准备命令成功:
node modules/llm-answer-reference-compare/assets/tool/dist/cli.js login \
--platform dknowc-chat \
--question "<用户原始问题>" \
--out <run目录>/capture
node modules/llm-answer-reference-compare/assets/tool/dist/cli.js login \
--platform doubao \
--question "<用户原始问题>" \
--out <run目录>/capture不得代替用户填写账号、密码、短信验证码或处理人机验证。登录会话只保存在 ~/.fact-check-x/browser-profiles,不得写入技能包、任务产物或报告。已有持久化会话可以减少用户登录操作,但不能跳过载体能力预检、前台执行和可见页面成功验证。
豆包未登录页面也可能显示输入框。不得用“找到输入框”代替登录确认:豆包右上角登录入口仍可见时,只提示并等待用户完成登录,禁止填入或提交核验问题;登录入口消失并检测到可提问界面后才开始采集。
node modules/llm-answer-reference-compare/assets/tool/dist/cli.js run \
--question "<用户原始问题>" \
--platform dknowc-chat \
--platform doubao \
--out <run目录>/capture \
--headed \
--interactive \
--timeout 180000 \
--retries 2默认使用 2 个平台,也允许重复 --platform 选择任意 N 个平台(N≥1)。N=1 执行单平台知识点结构化与权威核验;N≥2 额外执行跨平台一致性和差异对比。
标准深知晓使用 dknowc-chat。深知晓(深度溯源)使用
dknowc-deep-research:采集器会在同一会话先等待普通回答完整生成,再点击
“深度研究”,接管新打开的可信溯源报告页,等待结果完整生成后独立保存为一个
平台;按钮缺失、报告页未打开或结果未完成均按采集失败处理。平台组合完全按用户输入决定,
正式支持的内置选项包括
dknowc-chat、dknowc-deep-research、doubao、yuanbao、deepseek、
qianwen;generic 仅供开发者适配和验证新网页,不属于正式支持平台。
不存在固定“五平台模式”或固定上限。所有选定平台必须
采集成功后才允许进入知识点对比。原始答案、知识点对比、权威证据核验与平台表现报告均按本次
实际平台集合动态展示。
保留完整原答案、原始 URL、引用标记、引用正文、截图和失败状态。不得摘要、改写或用搜索结果替换原始来源。
当深知平台所附来源只有标题或截断摘要时,采集阶段用可信搜索 return_full_content=true 补全与该标题或原始 URL 匹配的同一材料全文;返回全文或段落直接用于判断,不访问源网址二次抓取正文。深知晓与深知晓(深度溯源)的可信搜索材料统一按“官方材料”处理,不以 .gov 域名或外链是否返回降级;返回的 源网址 仅作回溯链接,未返回时保留深知收录页且不伪造外链。其他平台的回答区上下文只用于现场存证,不得当作链接原文;已绑定 PDF 经可信搜索与直接提取仍无正文时必须失败关闭并进入 OCR 或 Computer Use,不能把采集缺口计为平台幻觉。
capture-recovery.json 并以非零状态退出。此时必须暂停流水线。capture-recovery.json;这已经是 Computer Use 接管信号,不是继续诊断浏览器环境的授权。capture-recovery.json 的 action 为 computer_use 时,有 Computer Use 的运行载体立即恢复同一平台,处理登录后的页面操作、地区选择、问题提交和回答完成等待;没有该能力时停在原始答案采集阶段,不得直接跳到知识点对比。capture-recovery.json.question,不得要求用户滚动到旧会话开头寻找或复制原问题。results.json 中所有指定平台均为 success、回答非空且 capture-recovery.json.status 不再是 required,才允许运行 prepare-comparison。capture-gate.json 是程序级硬门禁;失败平台、空回答和初始化提示都无法进入知识点对比、权威核验或最终报告。python3 scripts/fact_check_x.py prepare-comparison \
--results <run目录>/capture/results.json \
--run-dir <run目录>命令成功后,必须立刻向用户发送一条独立的 原始答案采集完成检查点,展示每个平台的采集状态、可回溯参考文献数量、无 URL 来源标签数量和耗时,并提供以下可点击产物:
capture/report.html:原始答案、参考文献和引用关系;capture/results.json:无损结构化采集结果;capture/report.md:可迁移文本报告;capture-gate.json:全部平台采集成功证明。必须使用命令返回的 deliverables[0].path 发送真正的 Markdown 文件链接,例如 [打开各方答案汇总](<返回路径>);禁止仅用反引号显示路径。该检查点必须在开始知识点对比前对用户可见。默认询问用户选择“继续下一步”“修正当前结果”或“到此结束并保留产物”,并等待选择;只有最初请求已明确授权完整自动跑完时才可不等待。不得只把路径藏在“运行命令”、折叠执行详情或最终总结中。
豆包等页面可能只显示来源名称而不暴露原文 URL。采集器会先尝试展开来源标签获取真实链接;仍无链接时写入 sourceMentions。此时必须表述为“0 条可回溯参考文献,N 个无 URL 来源标签”,不得说“页面没有来源”,也不得把标签伪造成参考文献。
读取 comparison-task.json,由当前承载智能体直接写 comparison-analysis.json。每个知识点只表达一个可核验事实变量;同一事实的不同数值必须对齐在同一点;只能使用原始答案采集阶段已经保存的来源判断来源忠实性,禁止联网补证。顶层必须填写 synthesisDraft,其 status 固定为 unverified,正文综合所有相关知识点并保留冲突、条件和缺口,basisKnowledgePointIds 只能引用当前知识点;它是“综合草案(未核验)”,不得写成权威最终答案。
原子性必须落实到每个平台的 claim,不能只把知识点标题写得宽泛。原回答一句话同时包含多个可独立判真的义务、条件、对象、数值或后果时,必须拆成多个知识点;各点的 claim 只保留当前事实,answerExcerpt 可以复用同一段原文。仅个别平台增加的实质事实也要单独成点,其他平台标为未覆盖,禁止把新增事实并入宽泛知识点后借用深知晓锚点免查。trustedAnchor 只能覆盖与深知晓权威结论相同的单一事实变量;超出该变量的主张必须成为无锚点知识点,交给后续一次可信搜索。
每个 covered=true 的 claim 必须填写 answerExcerpt:它必须是原始 answerMarkdown 的连续子串,并覆盖当前原子主张。载体负责知识点、主张和原回答片段的语义判断;程序负责从已捕获来源中校验脚标、重建可定位证据摘录、归一化引用方式,并自动生成合格的深知晓可信锚点。
answerExcerpt 内时列入 citedReferenceIndexes;当前主张已有局部脚标后,答案后段或回答级官方来源不得反向抬高它。citedReferenceIndexes 或 answerLevelReferenceIndexes。程序会在对应 capturedText 中定位支持当前主张的原文,并归一化为回答级语义溯源;定位失败时标记为分析信息不足,继续交给权威证据阶段裁决。local、declared_global、answer_level_semantic 或 none,最终报告分别外显为“逐段溯源 / 无对应的清单 / 全文语义溯源 / 未建立溯源”。缺少可定位证据时保守降级,不得默认为官方依据。python3 scripts/fact_check_x.py complete-comparison \
--results <run目录>/capture/results.json \
--run-dir <run目录>若命令返回 fact-check-x/product-truth@1 契约错误,必须停留在知识点对比阶段:
按错误路径补齐或重写 comparison-analysis.json,随后重新执行
complete-comparison,最多自动修复 2 次。两次后仍失败则明确报告知识点对比
阻断及缺失字段,禁止进入 prepare-authority,也禁止静默结束任务。
命令成功后,必须立刻向用户发送一条独立的 各方答案聚合完成检查点,展示知识点数量、分析信息不足数量和“综合草案(未核验)”。必须使用 deliverables[0].path 发送 [打开各方答案聚合(未核验)](<返回路径>),不能只写 comparison.html 或把表格补在最终答复中。默认询问用户选择“继续下一步”“修正当前结果”或“到此结束并保留产物”,并等待选择;只有最初请求已明确授权完整自动跑完时才可不等待。
可信搜索使用跨载体共享的本机配置。每次进入权威核验前先自动检查;已有有效 Key 时直接继续,不打开登录页,也不要求用户重复配置。首次缺少 Key 时,向用户说明“将打开深知 MaaS 页面,您只需完成登录,后续由技能自动配置”,随后立即执行命令返回的 configuration.command。配置组件会使用包内 Playwright 打开系统浏览器并等待用户本人完成短信、密码、验证码或人机验证;登录成功后自动读取已有完整 Key,没有可复用 Key 时创建名称为 Fact-Check-X 的专用 Key,验证后写入 ~/.fact-check-x/credentials/trusted-search-key。该配置供 Codex、Claude Code、WorkBuddy 等载体共同复用。
用户不需要查找、复制、粘贴或回复 Key,也不需要编辑 shell 文件、执行环境变量命令或回复“已配置”。配置命令成功后自动重新执行 prepare-authority 并继续。若用户在对话中粘贴疑似密钥,仍必须视为已经泄露:不得复制到命令、脚本、日志、配置或报告中,不得用该值继续执行;立即提示用户在 MaaS 控制台吊销,然后重新运行自动配置。
只有可信搜索返回 401/403 时才把现有 Key 视为失效并进入自动配置。超时、断网或服务异常必须保留现有配置,由搜索组件自动重试;重试后仍失败则停留在当前技术阶段并返回明确错误,不得要求用户重新登录,也不得转成事实复核任务。
python3 scripts/fact_check_x.py prepare-authority --run-dir <run目录>
python3 scripts/fact_check_x.py search-authority --run-dir <run目录> --max-workers 12prepare-authority 会先统计必须调用可信搜索的知识点。只要存在非免查知识点且本机共享配置与既有环境都没有可用 Key,命令就返回 status=configuration_required、action=configure_trusted_search、configuration.command 和可直接展示的 userPrompt,并以非零状态退出。此时先向用户展示简短登录提示,再立即前台执行返回的配置命令:
部分知识点还需要调用可信搜索。首次使用将打开深知智能 MaaS 平台,您只需完成登录;技能会自动获取或创建 Fact-Check-X 专用 Key,验证并保存后继续。本机已有配置时会直接跳过。
配置组件返回 status=configured 或 status=already_configured 后,自动重新执行 prepare-authority;只有返回 status=prepared 才能继续搜索。登录期间保持浏览器与命令运行,用户完成登录后自动续接。不得自行决定“直接基于深知晓官方来源裁决”,不得以普通联网搜索代替可信搜索,也不得让用户在对话中发送密钥。
comparison-gate.json 会锁定采集结果、comparison-analysis.json 与 comparison.json 的摘要;authority-gate.json 会继续锁定 request、evidence、assessment 和 result 的精确文件集合及摘要。后续发现文件被修改、ID 不一致、额外/陈旧 result、缺失 assessment(已取得证据的知识点)或手工结果时必须拒绝继续。只有 search-authority 正常完成并把门禁更新为 searched 后,才能写 assessment、运行 finalize-authority 或生成最终报告。禁止手工伪造 evidence、result,禁止通过修改中间 JSON 消除证据不足记录,禁止在缺钥时宣称“核心事实核验已完成”。
逐个读取 authority/requests 和 authority/evidence,由当前运行载体直接把裁决写入 authority/assessments/<知识点ID>.json。已有合格深知晓官方锚点时必须免查;否则每个知识点只调用一次可信搜索。多个知识点独立并发,不上传无关回答全文。
免查模式下,trustedAnchor.officialAnswer 是当前知识点的权威结论,锚点证据列表用于来源追溯,不要求每条标题或截断摘录逐字复述该结论。平台主张与 officialAnswer 语义一致,或可由其直接推出时,必须裁决为 supported 并引用当前锚点中的有效证据 ID。只有平台主张增加了 officialAnswer 和锚点均不能支持的实质事实,或者事实确实无法判定时,才使用 insufficient。不得仅因锚点摘录较短,就把与 officialAnswer 同义的主张判为证据不足。
知识点对比阶段的引用忠实性和本阶段的事实正确性必须分别保留:平台自己的引用不支持其主张时,继续记录 faithfulness=insufficient;若该主张经权威结论证实,本阶段仍裁决为 supported,最终分类由程序据此形成 coincidental,不能用引用缺口代替事实裁决。
每个裁决文件必须严格使用以下结构;平台键必须与 request 中已覆盖的平台 ID 一致,evidenceIds 只能引用当前 evidence 文件中的证据 ID:
{
"authoritativeFinding": "由权威证据支持的知识点结论",
"verdicts": {
"dknowc-chat": {
"verdict": "supported",
"reason": "主张与 E1 一致",
"evidenceIds": ["E1"]
},
"doubao": {
"verdict": "contradicted",
"reason": "主张与 E1 冲突",
"evidenceIds": ["E1"]
}
}
}verdict 只能是 supported、contradicted 或 insufficient。禁止使用顶层 verdict、officialAnswer 或 platformAssessment 代替规定字段。结构错误、证据 ID 不存在、平台裁决缺失都会被程序拒绝,必须修正后重新执行。
python3 scripts/fact_check_x.py finalize-authority --run-dir <run目录>finalize-authority 完成裁决后会独立生成 verification.json 和
03-authority-report.html。证据充分的知识点进入 finalAnswer;证据不足的知识点进入 evidenceGaps,从确定答案和准确率分母中排除,但命令仍返回 status=completed 并允许继续生成第四步。必须立即发送 权威证据核验检查点,使用
deliverables[0].path 提供 [打开全知晓“完美答案”](<返回路径>);报告顶部展示“完美答案(已核验,可权威溯源)”,无法证实也无法证伪的内容在下方单列供参考,并保留逐知识点证据、证据边界和平台裁决。默认询问用户
选择“继续下一步”“修正当前结果”或“到此结束并保留产物”,并等待选择;只有
最初请求已明确授权完整自动跑完时才可不等待。
用户确认后再生成平台表现评估与完整报告包:
python3 scripts/fact_check_x.py deliver \
--results <run目录>/capture/results.json \
--run-dir <run目录>finalize-authority 或 deliver 返回非零状态时,必须按技术错误修正或重试当前阶段,不得绕过门禁。技术流程成功但存在 evidenceGaps 时继续交付第四步,并明确这些项目不进入确定答案和准确率分母。
可信搜索返回成功但当前知识点 no_evidence 时,归类为 insufficient_evidence。单次检索未返回材料不等于官方明确否定,更不等于平台主张“编造”;程序必须完成本知识点的证据边界记录,并继续后续报告。
最终必须交付:
05-complete-report-package.zip:可直接发送给他人的完整报告包,内含四份 HTML、原始 JSON、截图、页面存证和逐知识点核验数据;01-capture-report.html:各方答案汇总;02-comparison-report.html:各方答案聚合(未核验);03-authority-report.html:全知晓“完美答案”、官方证据、证据边界和各平台裁决;04-final-report.html:按本次动态平台集合生成的各方答案测评报告;capture/results.json:原始回答、参考文献、原始 URL、引用标记和现场存证索引;capture/capture-recovery.json:在线采集时记录 Computer Use 恢复状态;capture/report.html:原始答案、参考文献与引用关系可视化报告;capture/report.md:原始答案与引用的可迁移文本报告;capture-gate.json:全部指定平台采集成功的程序级门禁证明;comparison-gate.json:知识点对比输入、分析与归一化结果未被后改的 provenance 证明;comparison.html:知识点结构化对比报告(未核验);verification.json、report.html、pipeline.json:权威核验数据、最终平台表现报告和全链路清单;最终对话必须使用 deliverables 返回路径,按“完整可分发报告包、各方答案汇总、各方答案聚合(未核验)、全知晓‘完美答案’、各方答案测评报告、全链路清单”给出实际状态。即使最终报告已经生成,也不得省略三个阶段报告;并允许用户确认完成或指定返回修正的阶段。
本机文件链接只用于当前运行会话内查看,禁止把 macOS 用户目录、Windows 本地盘符路径或本机文件协议地址描述成群成员、客户或共享链接访问者可以打开的对外交付地址。需要对外发送或转交时:
05-complete-report-package.zip 作为真实文件附件上传到目标会话、群聊、邮件或云盘;01-capture-report.html、02-comparison-report.html、03-authority-report.html 和 04-final-report.html;05-complete-report-package.zip 内部必须使用相对路径,解压后可直接打开四份报告,不得包含可执行机器的绝对路径。仅当知识点没有合格深知晓锚点且本机尚无共享配置时,统一入口会自动返回并执行:
python3 scripts/trusted_search_config.py configure命令会打开深知 MaaS 登录页。用户只完成登录,技能自动获取或创建专用 Key、 验证并保存;后续任何承载端均直接复用。Key 禁止写入对话、命令输出、报告、 请求文件或技能包。
trustedAnchor.eligible=true;深度溯源不继承普通深知晓的结果。© aiskillstore, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 118 other files (scripts, references) in skills/asi2030/fact-check-x-complete of aiskillstore/marketplace.
Open the folder on GitHubat commit 44923f3
Fact Check X Complete next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Fact Check X Complete this skillaiskillstore/marketplace | 433 | — | ~3.8k | Automated safety check: Pass | Apache-2.0 | |
| Researchzhongkaifu/TensorSharp | 568 | — | ~2.3k | Automated safety check: Warn | BSD-3-Clause | |
| Z Liang Wenfeng Grounded Voicetjxj/z-skills | 548 | — | ~1.4k | Automated safety check: Pass | None | |
| Dingo VerifyMigoXLab/dingo | 757 | — | ~741 | Automated safety check: Notes | Apache-2.0 | |
| Perplexity Web Searchdavila7/claude-code-templates | 33k | 11 repos | ~3.5k | Automated safety check: Notes | MIT | |
| Citation Verification GuideGalaxy-Dawn/claude-scholar | 5.7k | 2 repos | ~1.9k | Automated safety check: Pass | MIT |
zhongkaifu/TensorSharp
A skill your agent uses for web searches and current information lookups, finding sources, fact-checking, researching questions, comparing sources, or summarising web pages.
tjxj/z-skills
A skill your agent uses whenever the user asks to talk with, roleplay, interview, quote, fact-check, or apply the reasoning of 梁文锋/Liang Wenfeng from the bundled May 20 investor-meeting transcript.
MigoXLab/dingo
A skill your agent uses when the user wants to fact-check an article or verify factual claims in a document.
davila7/claude-code-templates
Runs web-grounded searches through Perplexity's Sonar models over OpenRouter for current events, recent literature and cited facts beyond the model's training cutoff.
Galaxy-Dawn/claude-scholar
Reference guidance for checking every citation in academic writing against canonical sources such as DOI, arXiv, CrossRef and Semantic Scholar, to catch fake or wrong references.
digoal/blog
三层审查模型,逐段逐句验证文章真伪、证据链与逻辑结构。Use when the user asks to fact-check, verify, audit, or evaluate the credibility of an article, essay, report, opinion piece, social-media post, or any written claim —…
aiskillstore/marketplace
Analyze codebase with tokei (fast line counts by language) and difft (semantic AST-aware diffs).
aiskillstore/marketplace
Process JSON with jq and YAML/TOML with yq. An agent skill from aiskillstore/marketplace.
aiskillstore/marketplace
Scans for project documentation files (AGENTS.md, CLAUDE.md, GEMINI.md, COPILOT.md, CURSOR.md, WARP.md, and 15+ other formats) and synthesizes guidance.
aiskillstore/marketplace
Modern file and content search using fd, ripgrep (rg), and fzf.
aiskillstore/marketplace
Modern find-and-replace using sd (simpler than sed) and batch replacement patterns.
aiskillstore/marketplace
Detects stale project plans and suggests session commands. An agent skill from aiskillstore/marketplace.
Works with
Categories
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、冲突与证据边界全程可追溯。基础采集与对比无需 API…. Fact Check X Complete is an agent skill from aiskillstore/marketplace.
Fact Check X Complete fits situations like: tasks that involve Fact-checking and source verification.
Run `npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a claude-code`. Or copy the skill folder (skills/asi2030/fact-check-x-complete in aiskillstore/marketplace) into .claude/skills/fact-check-x-complete in your project. Claude Code loads it when a task matches its description.
Run `npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a codex`. Or copy the skill folder (skills/asi2030/fact-check-x-complete in aiskillstore/marketplace) into .agents/skills/fact-check-x-complete in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add aiskillstore/marketplace --skill fact-check-x-complete -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/fact-check-x-complete, .gemini/skills/fact-check-x-complete, .github/skills/fact-check-x-complete and .opencode/skills/fact-check-x-complete in your project.
Going by SKILL.md and its folder, Fact Check X Complete needs Python for the scripts in its folder and the command-line tools its instructions call (python3, node and npm). Our summary lists: Python 3.
SKILL.md names 1 domain. As links in the text: raw.githubusercontent.com. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.
Fact Check X Complete is published under the Apache-2.0 licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.
About 3.8k tokens (SKILL.md is roughly 15k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 6.4k tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Fact Check X Complete: Research (zhongkaifu/TensorSharp, 568 stars), Z Liang Wenfeng Grounded Voice (tjxj/z-skills, 548 stars), Dingo Verify (MigoXLab/dingo, 757 stars) and Perplexity Web Search (davila7/claude-code-templates, 33k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
aiskillstore (a GitHub organization) maintains it in aiskillstore/marketplace, which has 433 GitHub stars. The repository holds 1,044 skills in this directory. The repository was last updated on October 10, 2026.
Source: aiskillstore/marketplace on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.