Fetches and learns from technical blogs to improve Claude Code usage.

Install

mkdir -p .claude/skills/cc-blog && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16763" && unzip -o skill.zip -d .claude/skills/cc-blog && rm skill.zip

Installs to .claude/skills/cc-blog

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Learn from blogs to improve Claude Code usage
45 charsno explicit “when” trigger
Intermediate

Key capabilities

  • read blogs from blogs.yaml
  • fetch latest articles using WebFetch
  • extract Claude Code related content
  • propose implementation steps
  • record effective content in CLAUDE.md
  • update state.yaml to mark articles as seen

How it works

The skill reads blog configurations, fetches new articles, compares them with the last check state, extracts Claude Code related content, and proposes implementation steps.

Inputs & outputs

You give it
blogs.yaml file and state.yaml file
You get back
formatted output of checked articles, key points, things to try, and implementation steps

When to use cc-blog

  • Monitor blogs
  • Learn AI patterns
  • Update internal docs

About this skill

Learn from Blogs to Improve Claude Code Usage

Read the latest articles from blogs defined in blogs.yaml and extract/practice content that can improve Claude Code usage.

Configuration

  • blogs.yaml - Blogs to monitor (in this skill directory)
  • blogs.example.yaml - Configuration example

State Tracking

  • state.schema.json - JSON Schema definition
  • state.yaml - Last check state (auto-created, gitignored)
  • state.example.yaml - Example with schema reference

Rules:

  • last_checked is Unix epoch seconds (UTC)
  • Schema comment (# yaml-language-server: ...) must be preserved on updates
  • Concurrent execution: last write wins (use atomic write: temp file + rename)
  • Orphan keys: remove entries from state.yaml if blog is deleted from blogs.yaml
  • Only update state for successfully fetched blogs

Recovery:

  • If state.yaml is corrupted, delete it and run /cc-blog again

Tasks

  1. Read blogs.yaml to get blog list
  2. Read state.yaml if exists (first run: treat all as new, create blogs: {})
    • If parse error: warn user and offer to delete and recreate
  3. Remove orphan keys from state (blogs not in blogs.yaml)
  4. For each blog:
    • Fetch latest articles using WebFetch
    • If fetch fails: report error and skip this blog (do not update state)
    • Compare with state.yaml.blogs.<name>:
      • If last_url matches: "No new articles"
      • If last_url differs: mark as [NEW]
  5. For NEW articles: Pick up content related to Claude Code / AI usage
  6. Extract specific things to try in your environment
  7. Propose implementation steps
  8. Record in CLAUDE.md or docs/claude-code-routine.md if effective
  9. Ask user: "Update state.yaml to mark these as seen?"
  10. If yes, update state.yaml (atomic write: write to temp file, then rename):
    • Preserve schema comment at top
    • Only update successfully fetched blogs
    • Set last_checked to current Unix epoch
    • Set last_url, last_title, last_published from newest article

Output Format

## Articles Checked

### Status
- laiso: Last check 1738540800 (3 days ago)
- Anthropic Engineering: Last check 1738540800 (3 days ago)

### [laiso]
- [NEW] [Article Title](URL) - 2026-02-01

### [Anthropic Engineering]
- No new articles

### Errors
- (none)

## Content Related to Claude Code Usage

### Key Points
- ...

### Things to Try in Your Environment
- ...

### Implementation Steps
1. ...
2. ...

## Actions

- [ ] Try now: ...
- [ ] Add to CLAUDE.md: ...
- [ ] Consider later: ...

---
Update state.yaml to mark as seen? (y/n)

Notes

  • If article is unrelated to Claude Code/AI usage, report "No related articles"
  • Don't force application
  • Only propose things that seem actually effective

When not to use it

  • when articles are unrelated to Claude Code or AI usage
  • when the user does not want to update state.yaml
  • when the user does not want to record content

Limitations

  • only processes blogs defined in blogs.yaml
  • only updates state for successfully fetched blogs
  • only proposes things that seem actually effective

How it compares

This skill automates the process of monitoring blogs for Claude Code patterns and suggesting actionable steps, rather than manually browsing and extracting information.

Compared to similar skills

cc-blog side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
cc-blog (this skill)06moNo flagsIntermediate
youtube-collector18moReviewIntermediate
ml-paper-writing486moReviewAdvanced
nsfc-research-content-writer44moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry