predict-content

Scores web articles against an LLM knowledge baseline to measure reading value.

Updated Jun 30, 2026

Search Tags:#webfetch #web content #content curation #article evaluation #reading triage #information scoring

Ready to run

npx skills add https://github.com/santoshkal/chezmoi --skill predict-content-santoshkal

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: predict-content
Source: https://github.com/santoshkal/chezmoi/tree/main/private_dot_config/opencode/skills/predict-content
Command: npx skills add https://github.com/santoshkal/chezmoi --skill predict-content-santoshkal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It helps you decide whether an article is worth reading by comparing its content against what an LLM already knows, so you can skip restatements of common knowledge and focus on genuinely novel material. ## Core Features & Use Cases - Prediction Baseline: Writes a prediction of the article's content from its title and domain before reading, establishing an honest knowledge baseline. - Three-Tier Scoring: Rates articles as `-` (restatement), `.` (useful nuance), or `+` (novel information or novel connection) with calibrated expectations. - Structured Verdict: Returns the delta, delta type, topic tags, a three-sided jury (read, skim, or skip), and follow-up search queries. - Use Case: Paste a URL from your reading backlog and get a scored verdict telling you whether to read it fully, skim specific sections, or delete it. ## Quick Start Ask the assistant to score the article at a given URL using the predict-content skill.

Frequently Asked Questions about predict-content

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I check if an article is worth reading before reading it?▼

Provide the article URL and the skill fetches it with WebFetch, writes a prediction of its content from the title and domain, then scores how much the actual article exceeds that prediction. You get a `-`, `.`, or `+` verdict with reasoning.

What do the article scores minus, dot, and plus mean?▼

A `-` means the article restates common knowledge, `.` means it adds useful nuance findable elsewhere, and `+` means it contains novel information or a novel connection you would lose by skipping it. Most articles are expected to score `.`.

Can this skill evaluate articles behind a paywall or login?▼

No. The skill relies on WebFetch to retrieve article content, so pages that block fetching, require authentication, or are paywalled cannot be fully evaluated. It works best on publicly accessible articles.

What is a novel connection versus novel information in scoring?▼

Novel information is primary experience, insider context, or non-public detail absent from the common record. A novel connection links separately known ideas in a way that changes how you reason or act, such as showing A causes B so you should do C differently.

What are the limitations of LLM-based article scoring?▼

The score depends on the model's existing knowledge of the topic, so niche subjects may inflate `+` ratings while expert readers may disagree. It also cannot verify factual accuracy of the article's claims.