kk6

Intent-Driven Development (IDD) に基づき、ADR（Architecture Decision Record）形式の Intentドキュメントを作成するスキル。ユーザーの「こういう機能を作りたい」「この設計判断をした」 という意図をWhy/Whatに整理し、How（実装詳細）の混入を検出・分離して、docs/adr/ に連番の ADRファイルを生成する。 このスキルは以下の場合に必ず使用すること： - ユーザーが /intent と入力したとき - 「ADRを書きたい」「意図を記録したい」「intentを作成」「設計判断を残したい」「ADR作って」 「決定を記録」「なぜこうしたか残しておきたい」などと言ったとき - ユーザーが新機能やアーキテクチャの判断について議論し、それを文書化したいと示唆したとき Do NOT trigger for: バグ修正の記録、ルーチンなリファクタリング、意図が自明な些細な変更。

Automatically generate PR descriptions according to project templates.

Structured 5W1H completeness checker for written communication. Analyzes text using a Who/What/When/Where/Why/How framework, produces a status table for each element, and provides concrete rewrite suggestions for missing information. ALWAYS use this skill when the user shares text and asks you to review it, check it, proofread it, or confirm it's ready to send — including progress reports, daily standups, meeting minutes, incident reports, handoff documents, status updates, client emails, Slack messages, PR descriptions, or any prose meant to inform others. Also trigger when the user says things like "これで大丈夫か", "情報足りてるか", "抜けてないか", "伝わるかな", "5W1H", "check if anything is missing", "is this clear enough", "review before sending", or asks you to check whether a written message has enough context for the reader. This skill adds a structured framework that generic reviewing lacks — do not skip it just because you could review text without it. Do NOT trigger for: code review, security review, grammar/style-only che

Review a draft text (message, report, PR description, commit message, etc.) from the reader's perspective and suggest improvements.

Safely add a new dotfile to chezmoi management, determining the correct prefix and committing the result.

Scan chezmoi source files for accidentally exposed secrets and verify sensitive files have the private_ prefix.

Inspect the current chezmoi sync state and report unapplied changes, unmanaged files, and whether it is safe to run chezmoi apply.

Skills to assist in debugging Python code

Maintain project documentation by extracting specifications from test cases and implementation code. Creates or updates README, CLI usage guides, TUI guides, and API references using tests as the source of truth. Use this skill when the user asks to update docs, check if docs are outdated, write documentation for a feature, generate a usage guide, or sync docs with code. Also trigger for phrases like "ドキュメントを更新して", "READMEが古くなってないか", "この機能の ドキュメントを書いて", "docsを最新にして", "使い方を書いて", "doc maintenance", "sync documentation", "generate docs from tests". This skill should also be considered when preparing a PR that changes CLI commands, TUI screens, or public APIs — documentation may need updating alongside the code.

Audit test coverage as living specification by comparing implementation against test cases. Detects specification gaps (public APIs without tests), contradictions (tests asserting behavior that differs from implementation), stale tests (testing removed/changed code), and missing edge cases. Use this skill when the user asks to check test coverage, verify spec coverage, audit tests, find contradictions between tests and code, or ensure tests serve as reliable specification. Also trigger for phrases like "are my tests up to date", "do tests match the implementation", "spec coverage", "test audit", "仕様カバレッジ", "テストと実装の矛盾", "テストの網羅性". Supports scoped checks (e.g., "check CLI spec coverage", "audit TUI tests") as well as full-project audits.

Review and minimally update Python project documentation after code changes. Use when an agent needs to check whether docstrings, inline comments, README files, or Markdown under docs/ still match the implementation, especially after modifying public APIs, behavior, configuration, side effects, constraints, examples, or user-facing workflows in a Python repository.