メニュー
Use when explicitly asked to walk through a plan or list of issues one-by-one, with the user deciding each via AskUserQuestion. Triggers only on a request to review/triage a plan interactively — NOT on any casual mention of a feature, fix, or bug.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Browse Bluesky content via API and firehose - search posts, fetch user activity, sample trending topics, read feeds and lists, analyze and categorize accounts. Supports authenticated access for personalized feeds. Use for Bluesky research, user monitoring, trend analysis, feed reading, firehose sampling, account categorization.
Reference documentation for invoking Codex through the wrapper scripts. This skill should be used when calling Codex, invoking Codex CLI, delegating tasks to Codex, running Codex in background, or managing Codex sessions. Covers codex-start.sh, codex-wait.sh, codex-get-session-id.sh, query file handling, and session management.
Use after implementation is complete and lint/tests pass to run a multi-agent review of uncommitted changes. Delegates to multiple perspectives (Codex, code reviewer, code architect, project-steward) checking for bugs, wiring/call-graph, dead code, and long-term project alignment.
This skill should be used when the user mentions "hugo basics", "hugo config", "hugo commands", "project structure", "hugo setup", "hugo.toml", "config.toml", "hugo serve", "hugo build", "content directory", "layouts directory", "static directory", "assets directory", "hugo configuration", "module mounts", "hugo environment", ".Site.Author", "author config", "site params", or any general Hugo static site generator questions. Provides foundational Hugo knowledge including project structure, configuration patterns, essential commands, and migration guidance for deprecated features.
This skill should be used when the user mentions "stryker", "mutation testing", "mutant", "mutation score", "survived mutant", "NoCoverage", "stryker disable", "mutation report", "equivalent mutant", "stryker incremental", "stryker cache", or any mutation testing of JavaScript/TypeScript code. Provides comprehensive Stryker mutation testing patterns, disable comment syntax, and strategies for handling survived mutants and NoCoverage results.
This skill should be used when the user mentions "write tests", "unit test", "fake timers", "mock timers", "test setTimeout", "test setInterval", "flaky tests", "test timing", "test concurrency", "mutation testing quality gate", "test quality", or any testing of JavaScript/TypeScript code. Provides mandatory standards for testing time-dependent code and mutation testing as a secondary quality gate.
| name | plan-review |
| description | Use when explicitly asked to walk through a plan or list of issues one-by-one, with the user deciding each via AskUserQuestion. Triggers only on a request to review/triage a plan interactively — NOT on any casual mention of a feature, fix, or bug. |
Review and discuss issues that need fixing or implementing, one at a time.
Identify issues to review from conversation context. Look at the current conversation for a plan, a list of issues, or any prior discussion of features, fixes, or bugs. Work from whatever is already in context — do NOT scan the codebase independently for issues.
For each issue, explain it clearly. Describe the issue as though the user doesn't know much about the code or the problem. Cover:
Use AskUserQuestion for each issue. Present the user with clear options for how to proceed — e.g., fix now, skip, defer, or take an alternative approach. Let the user drive the decision.
Ask clarifying questions freely. If anything is unclear at any point — about the codebase, the user's intent, or the best approach — use AskUserQuestion immediately rather than guessing.
Proceed through all issues sequentially. Don't rush or batch. One issue at a time, fully resolved before moving to the next.
After all issues are decided, summarize, record, and delegate. Recap the decisions made, then record the chosen actions into the task list with TaskCreate — capturing any sequential (→) or parallel (||) shape between them. Then DELEGATE the chosen fixes to sub-agents per the orchestrator rules in ~/.claude/CLAUDE.md. This skill itself does NOT start writing code — it routes the decided work to sub-agents.
