rss-catchup（RSS 追赶）

Catch up on RSS feeds by auto-summarizing new articles.

Prerequisites

Requires feedparser:

pip install feedparser

Helper Script

Use scripts/rss_helper.py for fetching data:

# List recent articles from a feed
python3 scripts/rss_helper.py feed URL [limit]

# Get article content (fetches full page)
python3 scripts/rss_helper.py article URL

Workflow

1. Load configuration

- Read feed list from references/feeds.json - Read last run timestamp from references/state.json - If first run: go back ~3 months - Otherwise: only articles since last run

1. For each enabled feed:

- Run: python3 scripts/rss_helper.py feed "FEED_URL" 20 - Filter to articles published since last run - Deduplicate against existing notes: grep -rl "media: {article_url}" "my-vault/07 Knowledge Base/Capture/Articles/" - If a note with that URL exists: skip (already processed, may be missing from state)

1. For each new article:

- If feed provides full content, use that - Otherwise, run: python3 scripts/rss_helper.py article "ARTICLE_URL" to fetch full text - Summarize the content into bullet points (3-8 based on length/density) - Create article note with status: Summarized - Extract discoveries (see below)

1. Create discovery notes:

- While summarizing, identify any products, services, frameworks, tools, libraries, or technologies mentioned that might be worth exploring - Search my-vault to check if a note already exists for each discovery - For new discoveries, create a note in my-vault/01 Inbox/ - Link the discovery note from the article's Related field

1. Update state

- Write current timestamp to references/state.json - Report summary of what was processed

Tagging

IMPORTANT: Read my-vault/09 System/Tag Index.md before processing to verify valid tags.

Tags MUST come from the canonical list - do not invent new tags. Common valid tags for this skill:

• #ai - LLMs, agents, prompting, AI tools
• #llm - Large Language Models, model comparisons
• #dev-tools - IDEs, Git tooling, developer productivity
• #python, #javascript, #typescript - language-specific
• #devops, #api, #databases - infrastructure topics
• #atlassian, #jira, #confluence - Atlassian products

Each feed in references/feeds.json has a tags array specifying default tags. Use these for article notes. Format: tags: ["tag1", "tag2"]

For discovery notes, choose tags based on what the discovery is (e.g., a Python library gets #python, an AI tool gets #ai).

Article Note Format

Create in: my-vault/07 Knowledge Base/Capture/Articles/[Feed Name]/[Title].md

Sanitize filenames: remove special characters, limit length to ~80 chars.

---
class: Article
media: https://example.com/article-url
publishDate: YYYY-MM-DD
status: Summarized
author: Author Name
reviewFrequency:
lastReviewedDate:
review:
aliases:
tags: ["tag1", "tag2"]
cssclasses:
archived:
---
Related:

## Summary

Capture the actual conclusions and insights - what would someone learn from reading this? Not topic labels or "this article discusses X" but the substance:

**Good:** "Multi-agent systems outperform single agents when context exceeds what fits in one prompt - Anthropic's research system with Opus 4 lead + Sonnet 4 subagents beat single-agent Opus 4 by 90.2%"

**Bad:** "Discusses multi-agent architectures and when to use them"

Aim for 4-8 substantive bullets that capture the key takeaways, conclusions, data points, and actionable insights.

## Discoveries

- [[Product Name]] - brief context from article
- (or "None" if nothing noteworthy)

## Why Read?

[One sentence on whether this seems worth actually reading in full]

Feed Config

Edit references/feeds.json:

{
  "feeds": [
    {
      "name": "Feed Display Name",
      "url": "https://example.com/feed",
      "folder": "Folder Name",
      "tags": ["tag1", "tag2"],
      "priority": "high",
      "enabled": true
    }
  ]
}

Tags should be from the canonical list in my-vault/09 System/Tag Index.md.

Discovery Note Format

Create in: my-vault/01 Inbox/[Name].md

---
class: Note
reviewFrequency:
lastReviewedDate:
review:
aliases:
tags: ["tag1", "tag2"]
cssclasses:
archived:
---
Up:
Related: [[Article Title]]

## What is it?

[One sentence description of the product/service/framework]

## Why look into it?

[Brief note on why it seemed interesting from the article context]

## Links

- [Official site or docs if mentioned]

What counts as a discovery:

• Products or services (SaaS tools, apps, platforms)
• Frameworks or libraries (programming, ML, etc.)
• Technologies or protocols
• Notable companies or projects
• Methodologies or techniques worth researching

Skip creating notes for:

• Well-known mainstream things (e.g., "Python", "AWS", "React")
• Generic concepts that don't warrant their own note
• Things already covered extensively in existing notes

Processing Tips

• Process one feed at a time and report progress
• For long articles, focus on main arguments/takeaways
• If article content can't be fetched, summarize from title/description
• Skip articles that already exist in my-vault
• Keep summaries concise - this is for deciding what to read, not replacing reading
• When extracting discoveries, be selective - only create notes for things genuinely worth exploring
• Deduplicate by URL - some articles may appear in multiple feeds

Path Handling

CRITICAL - Never escape spaces with backslashes:

• Use paths exactly as shown: my-vault/07 Knowledge Base/... (with literal spaces)
• The Write tool handles spaces correctly - backslash escaping creates literal \ characters in directory names
• When using Bash commands, wrap paths in double quotes: "my-vault/07 Knowledge Base/..."