Article Lens / 文章透视镜

Article Lens is a Codex skill for critical reading. It does not summarize an article for the reader. It helps the reader decide whether an article is worth reading, and see how the article persuades through facts, claims, evidence, reasoning, emotion, and intent.

文章透视镜不是“文章总结器”，而是“阅读判断力教练”：先读懂，再看清，最后判断。

What It Does

• Judges whether an article deserves close reading, selective reading, skimming, or skipping.
• Identifies the author's persuasion goal instead of simply summarizing the article.
• Separates facts, claims, evidence, reasoning, emotion, and marketing intent.
• Flags likely logical fallacies with cautious wording.
• Gives writers practical feedback on what to learn, what to avoid, and how to improve.
• Supports ad hoc analysis and optional batch inbox workflows.

Rating Scale

• A: Worth close reading. Dense information, clear claims, strong evidence, complete reasoning, clear boundaries, and awareness of counterexamples.
• B: Worth selective reading. Has useful claims, cases, or insight, but evidence, reasoning, or boundaries are not fully complete.
• C: Skim only. Clear stance, stronger emotion, weaker facts or evidence. Useful for understanding one position, not as a judgment basis.
• D: Skip. Heavy emotional manipulation, weak evidence, severe logic jumps, obvious marketing, attack, or factional framing.

Install

Copy the skill folder into your Codex skills directory:

cp -R skills/article-lens ~/.codex/skills/article-lens

Restart Codex or reload skills. Then ask:

Use Article Lens to analyze this article:
<paste URL or article text>

If you already have a private article-lens skill, install this public version under another name, such as:

cp -R skills/article-lens ~/.codex/skills/article-lens-public

Then rename the name field in SKILL.md to article-lens-public.

Usage

1. Ad Hoc Analysis

Paste an article URL or article text. The skill outputs a short report in the required format.

Default mode is quick mode, under about 800 Chinese characters. Use words like “深度分析”, “详细拆解”, or “做成报告” to request a longer analysis.

2. Batch Inbox

This repo includes an optional storage schema for tools such as Lark Base, Airtable, Notion database, Google Sheets, or a local CSV/SQLite queue.

The skill itself does not require any storage provider. Batch mode works when you provide an integration that can:

• collect article URLs into an inbox;
• mark records as pending or analyzed;
• write the analysis fields back into the same record.

See Storage Schema.

3. Browser Clipper

A browser clipper can be built on top of the storage schema. The recommended design is:

1. A browser extension collects the current page title, URL, author if available, selected text, tags, and notes.
2. The extension sends the payload to a local service or your own backend.
3. The service writes a pending record to your chosen inbox.
4. Later, Codex runs Article Lens batch analysis and updates the records.

Do not put private API tokens directly inside a browser extension.

Privacy

This public repo intentionally does not include:

• private Lark/Feishu base tokens;
• local filesystem paths;
• user article history;
• browser extension secrets;
• personal prompts or private analysis records.

If you add integrations, keep credentials in local config files or environment variables that are ignored by git.

Files

skills/article-lens/SKILL.md              Core Codex skill
skills/article-lens/agents/openai.yaml    Optional skill metadata
skills/article-lens/references/           Storage and batch mode references
examples/                                 Example input and output
integrations/                             Optional integration notes and templates

License

MIT. See LICENSE.