shimo4228
Community@shimo4228 · Japan
I run AI agents and publish everything I learn, failures included, from Claude Code harness engineering to an experiment running fully local on one M1 Mac.
Agent Skills by shimo4228
Showing 40 vetted skills indexed across 1 GitHub repositories.
xcode-package-swift-misidentification
Diagnose Xcode misidentification between Package.swift and .xcodeproj directories.
service-layer-extraction
Extract business logic from Typer/Click CLI modules into a testable service layer.
skill-stocktake
Audit Claude skills and commands with a repeatable stocktake workflow.
python-immutable-accumulator
Create immutable Python accumulators using frozen dataclasses and tuples.
python-module-to-package-refactor
Update unittest.mock.patch targets to runtime lookup paths during Python module-to-package refactors.
prh-hyphen-regex-escape
Fix hyphen-related regex escapes in prh dictionaries for Node.js 20+.
textual-tui-pipeline-interception
Insert a Textual UI review step into CLI pipelines before final write.
swift-actor-persistence
Provide thread-safe local data persistence with actor-based repositories and in-memory caching.
zenn-qiita-crosspost-workflow
Convert Zenn articles to Qiita syntax and post via publish.py CLI.
tech-writing-patterns
Cross-post technical articles from Zenn to Qiita with format and tone alignment.
zenn-markdownlint-config
Configure markdownlint-cli2 with Zenn-specific rule overrides for articles and books.
claude-code-mcp-manual-install
Edit ~/.claude.json mcpServers entries with jq for HTTP and stdio servers.
content-hash-cache-pattern
Cache expensive file processing results using SHA-256 content hashes.
immutable-model-updates
Create immutable Swift model updates via factory methods for thread-safe state.
ai-era-architecture-principles
Determine whether to use a large framework or custom implementation for Claude Code projects.
xcode-pbxproj-file-registration
Register new Swift files into four Xcode pbxproj sections.
parallel-subagent-batch-merge
Generate and consolidate 50+ structured items via parallel Claude Code subagents.
cross-source-fact-verification
Cross-validate factual claims in drafts against multiple independent sources.
regex-vs-llm-structured-text
Parse structured text with regex-first extraction and LLM fallback.
cost-aware-llm-pipeline
Implements immutable logging and automatic rerouting for cost-effective LLM inference.
keyword-based-llm-eval
Evaluate structured LLM outputs with keyword matching and F1 scoring.
zenn-textlint-workarounds
Document workarounds for Zenn article linting false positives from textlint and prh.
mock-friendly-api-layering
Prevent internal routing parameters from leaking into public API wrappers.
directory-structure-enforcement-hooks
Enforce file placement rules with three-layer Claude Code hooks.