跳到正文
cccskills
免费在 GitHub 公开

multimodal-evidence

Extract and verify text, numbers, objects, and sequences from images, documents, and video. Record where each finding appears, check exact labels and values, and build answers whose supporting evidence can be traced back to the input.

中文简介正在准备中。当前显示英语简介。

原文查看原文说明

Extract and verify evidence from images, documents, and video without redundant inspection

查看安装方法

适用场景

  • Extracting and checking numbers in documents
  • Tracing video events to timestamps
  • Reading and verifying labels in images
  • Checking evidence coverage before answering

包含的文件(1)

  • SKILL.md1.9 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

When to Use

Use when the answer or requested artifact depends on visual, temporal, tabular, or textual evidence contained in images, documents, or video.

Procedure

  1. Identify the evidence required: objects, text, values, ordering, timestamps, labels, or visual relationships.
  2. Inspect the whole input or a broad representative sample first to establish structure and likely evidence locations.
  3. Narrow to relevant pages, frames, regions, or time intervals and record observations with their locations.
  4. Use the format's native parser for exact text and numbers: for example python-docx or ZIP/XML inspection for DOCX, pdftotext or a PDF library for PDF, spreadsheet readers for XLSX, and OCR only when the source is image-based. Do not search binary office files with plain grep or cat.
  5. Resolve conflicts with one targeted reinspection at better scale or a nearby frame rather than repeating the same crop.
  6. Build the answer or artifact from the evidence ledger and perform a final coverage check against every requested item.

Pitfalls

  • Do not infer unseen content from filenames, surrounding text, or a single thumbnail.
  • Do not repeatedly inspect nearly identical regions without a new hypothesis.
  • Do not trust OCR blindly for small labels, punctuation, or numeric values.
  • Do not finalize before checking that every requested item has supporting evidence.

Verification

  • Each factual output can be traced to a page, frame, region, or timestamp.
  • Exact labels and numbers were visually checked after extraction.
  • The final response or artifact covers all requested evidence categories.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Turn scattered messages, meeting notes, and documents into sourced action lists and status briefs. The skill reconciles updates, separates facts from tentative inferences, and highlights decisions, dependencies, risks, and unanswered questions.

中文简介正在准备中。当前显示英语简介。

  • Build a sourced action list from meeting notes
  • Combine messages and documents into a status brief
  • Assess feasibility using dependencies and blockers
odysseus-dev/odysseus9.4万2026年10月17日 更新

Create a complete first version of a requested file or other deliverable early, then refine it and check its format, required content, location, and usability before reporting the work as finished.

中文简介正在准备中。当前显示英语简介。

  • Creating a report with all required sections
  • Saving a deliverable in a specified location
  • Checking that a generated file opens correctly
odysseus-dev/odysseus9.4万2026年10月17日 更新

Find official documentation for a library, framework, or API and apply it to the task at hand. Check the relevant version, distinguish documented facts from assumptions, and verify code changes with a focused test or concrete command.

中文简介正在准备中。当前显示英语简介。

  • Checking how an API or option works
  • Implementing behavior for a specific version
  • Applying official guidance to an existing project
odysseus-dev/odysseus9.4万2026年10月17日 更新

odysseus

免费

Use Claude Code to manage Odysseus tasks, email, calendar events, stored facts, and documents through a permission-scoped API. You can also start, inspect logs for, and stop Cookbook model servers using the access granted by your token.

中文简介正在准备中。当前显示英语简介。

  • Creating a task with a timed reminder
  • Reading email and preparing a reply draft
  • Saving calendar events, preferences, and documents
odysseus-dev/odysseus9.4万2026年10月17日 更新

odysseus

免费

Use Codex to manage Odysseus todos, calendars, memory, and documents, and read or draft email. Through its permission-scoped API, the skill also supports reminders and launching, inspecting, or stopping Cookbook model servers.

中文简介正在准备中。当前显示英语简介。

  • Create a timed reminder as a todo
  • Check calendar events and add appointments
  • Read email and prepare an editable reply draft
odysseus-dev/odysseus9.4万2026年10月17日 更新

Prepare reviewable updates for clients, partners, or leadership from internal messages and documents. The skill checks current figures, dates, and commitments, flags unresolved discrepancies, and excludes private internal details from the draft.

中文简介正在准备中。当前显示英语简介。

  • Draft a client progress update from internal records
  • Check corrected figures before preparing a partner update
  • Prepare a leadership brief with uncertainties flagged
odysseus-dev/odysseus9.4万2026年10月17日 更新

查看 odysseus-dev 的全部技能

报告此技能的问题