This skill automatically fetches content from Mastodon and Linkding, then composes it into a well-formatted Jekyll-style blog post with proper voice, tone, and narrative structure. No more copy-paste dumps—get readable, conversational weeknotes that sound like you.

Features

Multi-source data fetching: Mastodon posts and Linkding bookmarks
Conversational composition: Claude reads your content and composes readable prose
Style matching: Optionally reference your past weeknotes to maintain consistent voice
Smart tagging: Automatically generates 3-7 contextually appropriate tags
Jekyll-ready output: YAML frontmatter with proper filename conventions
Cross-platform: Supports macOS (ARM64/Intel) and Linux (AMD64)

Quick Start

Installation

Install this skill as a Claude Code marketplace:

# Add to your Claude config
# ~/.claude/config/settings.json
{
  "plugins": [
    "/path/to/lmorchard-agent-skills-private"
  ]
}

Restart Claude Code to load the skill.

First-Time Setup

The first time you use the skill, Claude will guide you through configuration:

User: Draft weeknotes for this week
Claude: I need to configure the skill first. I'll need:
  - Your Mastodon server URL and access token
  - Your Linkding instance URL and API token
  - (Optional) URL to your past weeknotes for style reference

Getting API credentials:

Mastodon: Settings → Development → New Application (read permissions)
Linkding: Settings → Integrations → Create Token

Basic Usage

# Default: last 7 days (rolling 7-day window)
Draft weeknotes for this week

# Specific date range
Create weeknotes from November 4-10

Claude will:

Fetch your Mastodon posts and Linkding bookmarks
Analyze the content for themes and topics
Compose conversational prose that sounds like you
Generate contextually appropriate tags
Save to your blog directory (if detected) or offer to save elsewhere

Configuration

After initial setup, your config lives in config/config.json:

{
  "mastodon": {
    "server": "https://your-instance.social",
    "token": "your-access-token"
  },
  "linkding": {
    "url": "https://your-linkding.com",
    "token": "your-api-token"
  },
  "weeknotes_archive": "https://yourblog.com/tag/weeknotes/"
}

To reconfigure:

./scripts/setup.sh

Project Structure

weeknotes-blog-post-composer/
├── SKILL.md              # Detailed documentation for Claude
├── README.md             # This file
├── bin/                  # Platform-specific Go CLI binaries
│   ├── darwin-arm64/
│   ├── darwin-amd64/
│   └── linux-amd64/
├── scripts/
│   ├── setup.sh          # First-time configuration
│   ├── fetch-sources.sh  # Fetch data from sources
│   ├── prepare-sources.py # Verify fetched data
│   └── download-binaries.sh # Update CLI binaries
├── config/
│   └── config.json       # API credentials (gitignored)
└── data/                 # Fetched markdown files (gitignored)

Data Sources

Currently supported:

Mastodon: Posts from specified date range
Linkding: Bookmarks from specified date range

The architecture supports adding additional data sources in the future.

Output Format

Generated blog posts include:

Jekyll YAML frontmatter with title, date, tags, and layout
Conversational prose composed from your content
Short contextual links (3-5 words) for readability
Inline images from Mastodon posts
3-7 tags including "weeknotes" plus contextual tags
Proper structure with TL;DR, main sections, Miscellanea, and conclusion

Filename Convention

When run from your blog directory, posts are saved to:

content/posts/{YYYY}/{YYYY-MM-DD-wWW}.md

Where:

{YYYY} = 4-digit year of start date
{YYYY-MM-DD} = Start date of the 7-day period
{wWW} = ISO week number (e.g., w16, w45)

Example: content/posts/2025/2025-11-07-w45.md

Manual Commands

You can run individual components if needed:

# Update binaries to latest releases
./scripts/download-binaries.sh

# Fetch data for specific date range
./scripts/fetch-sources.sh --start 2025-11-01 --end 2025-11-07

# Verify fetched data
./scripts/prepare-sources.py

Security

API credentials stored in config/config.json with 600 permissions
Config file is gitignored
Temporary config files cleaned up after use
Fetched data is gitignored

Documentation

For detailed documentation on how the skill works and how Claude uses it, see SKILL.md.

Requirements

Claude Code: Latest version
Bash: For shell scripts
Python 3: For Python scripts
curl: For API testing (during setup)
jq (optional): For better config parsing

Binaries for Mastodon and Linkding fetching are included for all supported platforms.

License

This is a personal skill for use with Claude Code. Binaries for mastodon-to-markdown and linkding-to-markdown are subject to their respective licenses.