Analyzes single or multiple GitHub sources (PR, commit, issue) and generates articles sharing practical technical insights. Matches the writing style of existing articles in the posts/
directory, written in Japanese as personal notes without addressing readers directly.
Source Analysis
Parse single or multiple sources from arguments
Fetch diffs and context using gh
command for each source
Automatically extract repository information from each source
Identify common themes, patterns, and technical insights across sources
Article Structure
Adjust focus based on specified direction
Technical challenges and solutions
Key implementation points and learnings
Code examples and best practices
Knowledge applicable to other developers
Content Development
Generate compelling title from implementation
Document problems solved and learnings
Include technology stack used
Add concrete implementation with code examples
Match writing style of existing articles in posts/
directory
Write as personal notes in Japanese without addressing readers
Verify technical terms and concepts using WebSearch/WebFetch for accuracy
Include reference URLs using markdown links or {{ embed URL }}
macro
Output Generation
Create article in Markdown format
Save to posts/
directory with filename $(date +%Y%m%d)-[slug].md
# Generate from single source
/blog https://github.com/owner/repo/pull/123
/blog https://github.com/owner/repo/commit/abc123def
# Generate from multiple sources
/blog https://github.com/owner/repo/pull/123 owner/repo#456 owner/repo@abc123def
/blog https://github.com/owner/repo/commit/abc123 https://github.com/owner/repo/commit/def456
# Multiple sources with specified direction
/blog performance optimization analysis \
https://github.com/owner/repo/pull/123 \
https://github.com/owner/repo/pull/124 \
https://github.com/owner/repo/commit/abc123def \
owner/repo#456
# Mix different repositories
/blog comparing different approaches owner1/repo1#123 owner2/repo2#456
Real Code: Always include implementation code
Context First: Explain why before how
Reproducible: Ensure examples are testable
Honest Reflection: Include failures and learnings
Practical Focus: Emphasize applicable knowledge
Version Awareness: Note specific versions used
Performance Data: Include metrics when relevant
Direction Consistency: Maintain specified direction throughout
Source Integration: Coherently integrate insights from multiple sources
# 1. Analyze writing style of recent articles in `posts/` directory
# Understand structure and style patterns of each article
# 2. Fetch content from GitHub URL (supports multiple sources)
# For PR
gh pr view [PR_URL] --json title,body,commits,files,additions,deletions,changedFiles
# For Commit
gh api repos/[owner]/[repo]/commits/[sha] --jq '{message,stats,files}'
# For Issue
gh issue view [ISSUE_URL] --json title,body,comments
# 3. Get detailed changes
gh pr diff [PR_URL]
# or
git show [commit_sha]
# or
gh api repos/[owner]/[repo]/compare/[base]...[head]
When generating articles, refer to recent article styles in posts/
directory.
Elements to reference:
Section structure
Writing style (personal notes in Japanese, use of past tense)
Usage of embed macro
Code example placement and explanation method
Avoid using colons before lists (use periods or other punctuation instead)
Generated articles include the following elements:
Frontmatter: title
, date
, tags
, rss_description
are required
+++
title = "RaycastでPLaMo翻訳を使う"
date = Date(2025, 06, 21)
tags = ["raycast"]
rss_description = "Script Command機能を使ってPLaMo翻訳CLIをRaycastから呼び出せるようにした"
+++
Body sections: Refer to structure of recent articles under posts/
External sources: Use markdown links and embed macro appropriately
Code examples: Excerpt from actual changes
Measurement data: Quantify performance and effectiveness
Parse GitHub URL patterns accurately (PR, Commit, Issue, etc.)
Use gh CLI for authenticated access
Dynamically learn writing style from latest ~5 articles
Auto-infer tags from changes
Excerpt and explain key parts when code changes exist
Structure multiple sources chronologically or logically
Organize performance data and measurements in tables
Honestly include failures and trial-and-error
Verify technical accuracy: Use WebSearch
/WebFetch
for unfamiliar terms, APIs, libraries, and implementation patterns
Cross-reference official documentation when explaining technical concepts
Citation practice: Always include referenced URLs in article body, not just as footnotes
Use embed macro for important references (official docs, GitHub issues, etc.)
Supported URL formats:
https://github.com/owner/repo/pull/123
https://github.com/owner/repo/commit/abc123def
https://github.com/owner/repo/issues/456
owner/repo#123
(PR/Issue)
owner/repo@abc123def
(Commit)
Generated article filenames:
posts/YYYYMMDD-[generated-slug-from-title].md
Date uses execution date from date +%Y%m%d
command