Deep Article Research

Given an article URL, fetch its content, extract core arguments, then use Gemini Deep Research for deep verification and analysis. Output a professional Chinese analysis report.

Requirements

Requirement	Details
GEMINI_API_KEY	Environment variable, required for Gemini Deep Research
agent-browser	For fetching JS-rendered article content (fallback: WebFetch)

Quick Start Workflow

Fetch article → Extract arguments → Gemini deep verification → Format report

Step 1: Fetch Article Content

Use agent-browser to fetch the full article content:

agent-browser --url "<article_url>" --action "extract the complete article text content including title, author, date, and full body text"

Save the raw content to tmp/article-deep-research/{slug}/article-raw.txt.

Fallback: If agent-browser is not available, use WebFetch tool to fetch the article.

If the article is behind a paywall or inaccessible, ask the user to paste the article text directly.

Generate {slug}: Create a short English identifier from the article title (e.g., china-ai-policy, fed-rate-decision). Use lowercase, hyphens, max 4 words.

Step 2: Extract Arguments (LLM Self-Execution)

Read and analyze the article following references/article-analysis-framework.md:

Identify the topic domain
Extract core arguments, evidence, logical chains, and implicit assumptions
Output structured analysis

Save to: tmp/article-deep-research/{slug}/argument-extraction-YYYY-MM-DD.md

Step 3: Gemini Deep Verification

python3.11 scripts/gemini_deep_research.py \
  --input <argument-extraction file path> \
  --prompt prompts/deep-verification-prompt.md \
  --output-dir tmp/article-deep-research/<slug> \
  --slug <slug> \
  --topic "<article topic keywords>"

Script paths are relative to this skill's directory. Use absolute paths when invoking:

--prompt: <skill_dir>/prompts/deep-verification-prompt.md
Script: <skill_dir>/scripts/gemini_deep_research.py

Where <skill_dir> is the directory containing this SKILL.md file.

Gemini will conduct web deep research to verify, refute, and supplement each argument, outputting a complete analysis report.

Output: tmp/article-deep-research/{slug}/{slug}-Analysis-Report-YYYY-MM-DD.md

Step 4: Format and Save

Format the report per references/markdown-formatter-rules.md, then save the final version.

Output Location

tmp/article-deep-research/
 +-- {slug}/
     +-- article-raw.txt                        # Raw article content
     +-- argument-extraction-YYYY-MM-DD.md      # LLM argument extraction
     +-- {slug}-Analysis-Report-YYYY-MM-DD.md   # Final analysis report

Reference Documents

Document	Purpose
article-analysis-framework.md	Argument extraction framework (LLM self-execution)
markdown-formatter-rules.md	Report formatting rules
prompts/deep-verification-prompt.md	Gemini deep verification prompt

Error Handling

Scenario	Action
Paywall / inaccessible article	Ask user to paste article text directly
`agent-browser` not installed	Fallback to `WebFetch`
`GEMINI_API_KEY` not set	Inform user: "Please set GEMINI_API_KEY environment variable"
Article too short	Still run full workflow (extraction + verification)

deep-article-research