kai kou
Community@kai-kou · osaka
2000年からWeb系のシステムエンジニアとして開発、運用、マネジメントを経験。2020年から副業で株式会社LABOTが運営するCODEGYMに講師として参加。現在はアライドアーキテクツでエンジニアマネージャー兼PjMとしてワイワイやってます。
Agent Skills by kai kou
Showing 22 vetted skills indexed across 1 GitHub repositories.
workflow-health-check
Audits GitHub PR, Issue, and pipeline health and applies safe automated fixes.
research-runner
Automates deep research generation with fallback engines, quality gates, and PR creation.
skill-audit
Audits Agent Skills directories for trigger conflicts, staleness, bloat, and broken references.
checkpoint
Creates git commits and Issue-comment state snapshots at pipeline phase boundaries.
audit-runner
Executes external Claude Code setup audit protocols and triages findings through discussion review.
self-improvement-loop
Automates cross-cutting project reviews, improvement issue triage, and backlog refinement on GitHub.
project-manager
Manages GitHub Issues, labels, milestones, and Projects V2 for task tracking workflows.
pr-review-watcher
Automates self-review, feedback resolution, and auto-merge for GitHub pull requests.
retro-try-handler
Detects, classifies, and implements open GitHub retrospective Try issues as pull requests.
code-review
Reviews pull request diffs with parallel fresh-context subagents and posts inline GitHub comments.
retrospective
Generates KPT retrospectives and records qualified Try items as GitHub Issues.
project-sync
Detects and resolves stale GitHub Issues, orphan PRs, and abandoned branches automatically.
typesafe-ai
Compose typed AI judgments and probabilities into application code using TypeSafe System One models.
skill-creator
Create, evaluate, and iteratively improve Claude skills with benchmark-driven testing.
claude-code-spec-sync
Tracks Claude Code changelog releases and updates project rules, skills, hooks, and settings accordingly.
self-reviewer
Detects and fixes code issues before pull request creation through layered self-review.
waiting-user-handler
Triage GitHub Issues labeled status:waiting-user and resolve or escalate them by category.
interactive-guide
Guides users step-by-step through manual operations that AI cannot execute directly.
apply-base
Synchronize shared rules, skills, and harness tooling from a base repository into the current repository.
discussion-review
Orchestrates adversarial multi-agent code reviews using named background agents and a shared whiteboard.
sprint-cycle-router
Routes cron-triggered sprint development cycles through a decision tree of GitHub automation branches.
cjk-markdown-fix-poc
Inserts half-width spaces around bold and inline code in CJK Markdown text.
Frequently Asked Questions About kai kou
FAQPage SchemaWhat tasks can I automate using kai-kou's skills?▼
You can automate GitHub Issue and milestone management, PR self-review and merge monitoring, stale issue and orphan branch cleanup, KPT retrospectives, deep research report generation, checkpoint snapshots, and cross-project self-improvement loops with Issue refinement and Epic grouping.
Who are these skills designed for?▼
They target software engineers and engineering managers running Claude Code-based development on GitHub repositories, especially teams practicing sprint cycles, TDD, and continuous improvement who want autonomous routine execution of reviews, triage, and repository hygiene.
How are the skills installed and executed?▼
Skills live under .claude/skills/ and .claude/commands/ in the repository and are invoked by slash commands, natural-language triggers, or scheduled routine slots. The apply-base skill syncs the shared claude-code-repository-base definitions into any target repository via git clone.
Are kai-kou's skills open source?▼
The typesafe-ai skill is explicitly MIT licensed. The remaining skills are published in public repositories such as claude-code-repository-base, which apply-base clones and applies, though individual license declarations are not stated in the manifest.
What prerequisites and dependencies do the skills require?▼
Skills require Claude Code with Bash, Read/Write, WebFetch/WebSearch, and GitHub MCP tools (issue, pull request, and actions operations). Some use Python3 scripts, the gh CLI, cron-style routine definitions, and companion skills like discussion-review and code-review.