Claude Skills by atman-33

Write a handoff document and launch the next agent in a split multiplexer pane (herdr or zellij), with new-window and paste-ready fallbacks when no multiplexer is in use.

Compact the current conversation into a handoff document for another agent to pick up.

Control herdr from inside it. Manage workspaces and tabs, split panes, spawn agents, read output, and wait for state changes — all via CLI commands that talk to the running herdr instance over a local unix socket. Use when running inside herdr (HERDR_ENV=1).

Launch a role-based team of Claude Code agents in separate multiplexer panes that coordinate on one task by messaging each other directly.

Install herdr (terminal-native agent multiplexer), set up its Claude Code and OpenCode integrations, and configure shell profiles to auto-launch herdr on startup (replacing any existing Zellij autostart block). Use when the user wants to install or configure herdr.

Install Zellij and configure shell profiles to auto-launch Zellij on startup (replacing any existing herdr autostart block). Use when the user wants to install or configure Zellij's shell autostart.

Use when the user explicitly asks to delegate work to a persistent sidekick/helper agent in its own pane that you can iterate with across multiple rounds (e.g. review -> fix -> re-review) without losing its context each time, or when another skill's instructions explicitly direct invoking sidekick-go. Do not use for a one-off review or one-shot subagent task; that's the Agent/Task tool's job.

Diagnose and fix `code .` failing to launch, or opening a disconnected Windows-side VS Code instead of a WSL-connected (Remote-WSL) window, in a herdr-spawned WSL shell.

Generate or refine a concise, effective CLAUDE.md for the current project following official best-practice guidelines. Use when the user wants to create, write, or update a project's CLAUDE.md.

Create a comprehensive, well-structured README.md for the current project. Use when the user wants to create, write, or refine a project's README.

Write user-facing release notes from the Conventional Commits history between two refs — grouped by impact, in plain language, with internal-only changes filtered out. Use when the user wants release notes, an announcement of what's new in a version, or a user-readable summary of changes since the last release (distinct from a developer changelog).

Compose a daily or weekly work log from git activity across the user's configured repositories and save it as Markdown (e.g. into an Obsidian vault). Use when the user asks for a work log, daily report, weekly summary, 日報/週報, or "what did I do today/this week".

Build a proof of a slide deck — one HTML file that steps through the slides in wireframe so structure, wording, and whether the text actually fits can be settled before any design is generated. Use when the user is preparing slides they will render in Claude Design, wants to iterate on a deck's contents, asks whether a slide is too full, or needs a paste-ready Claude Design prompt written from a settled structure.

Generate a rich, self-contained HTML document instead of a long Markdown file — for specs & plans, PR/code-review explainers, design mockups & prototypes, research/status reports, or throwaway editing UIs. Use when the user asks to write a "report", wants a PR or piece of code explained visually, wants to compare design options, needs a research or status writeup, wants a diagram/flowchart, or when a Markdown plan/spec is getting too long to read comfortably.

Post a message to a Slack channel, resolving channel/user names to IDs and adding an optional @mention or thread reply. Use when the user wants to post, share, or notify something in Slack, or when another skill needs to deliver its output to a Slack channel.

Turn a proposal into a decision-driven deck brief plus a ready-to-paste Claude Design prompt. Use when the user is preparing to pitch, review, or get sign-off on a proposal with their team, or asks how to structure the explanation so the meeting does not drift.

Use when building or improving a Claude Code slash command in a project's own `.claude/commands/`: $ARGUMENTS handling, file/bash injection, agent invocation, multi-agent orchestration, or command naming.

Grill the user relentlessly about a plan or design. Use when the user wants to stress-test a plan before building, or uses any 'grill' trigger phrases.

Install a Claude Code skill from a GitHub skills repository URL.

Create, back up, and restore Claude Code Desktop local routines (scheduled tasks) so they survive a PC change. Use when the user wants to set up a recurring or one-off Desktop routine, back up their routines, or restore routines on a new machine.

Reference for writing and editing skills well — the vocabulary and principles that make a skill predictable.

Capture a reusable, non-obvious project insight into the rules home where it will auto-inject in future sessions. Use when investigation or implementation surfaces knowledge worth persisting (a gotcha, a convention, the "why" behind a decision), or when the user asks to remember a project learning.

Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.

Commit the current changes with a Conventional Commits message. Use when the user asks to commit, or when a workflow (e.g. develop-small-feature) reaches its commit step after implementation is verified.

Record an architecture/design decision as an ADR (Architecture Decision Record) in the target repository's docs/adr/. Use when the user wants to document a design decision, its alternatives and consequences, or says "record this decision" / "write an ADR".

Creates a new feature branch from main inside the intended repository after confirming the branch target and repository context. Use when the user wants to start a new feature branch, asks to create a branch for upcoming work, mentions a branch name such as feature/<name>, or another workflow has already resolved the target repository.

Generate a self-contained HTML manual-testing guide for an implemented change — environment setup, a visual test-flow diagram, step-by-step scenarios with expected results, and interactive pass/fail checkboxes with a Markdown results export. Use when the user wants to verify an implementation by hand, asks "how do I test this", or wants a verification document for a feature/branch/PR.

Generate a self-contained HTML onboarding tour of a repository — architecture diagram, directory map, key flow walkthroughs, and a recommended reading order — for new team members. Use when the user wants onboarding material, asks for a codebase overview/architecture tour, or wants to explain a repo to someone new.

Analyzes git changes, drafts localized PR titles and bodies, and assists with creating or updating GitHub pull requests for the active target project repository. Use when working from the harness workspace (e.g. the workhub vault) and the user wants to create a PR, review branch changes, draft or update a PR description, or check whether a branch is ready for review.

Generate a self-contained HTML code-review guide for the current changes — big-picture overview, architecture diagram of affected modules, a file/class responsibility map, and an annotated change walkthrough — so a reviewer can grasp the whole implementation before diving into the diff. Use when the user wants review material for a branch/PR, asks to "explain the changes for review", or wants a visual summary of what was implemented where.

Implement a small, well-scoped feature or fix end-to-end from spec to PR — reuse or create a feature branch, implement test-first, run static checks, get user verification, commit, then open a PR to main.

A relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

Investigate a reported bug to its root cause — reproduce, trace the code path, identify the defect with evidence, assess blast radius, and propose fix candidates without changing code. Use when the user reports or pastes a bug/defect/unexpected behavior and wants it diagnosed, or asks "why is X happening".

Switch the openspecPath in .claude/project-context.json by picking one of the registered projects from a menu, instead of hand-editing the absolute path.

Run all engineering plugin setup steps in sequence — set up OpenSpec, scaffold project-context.json, and set up rules-ex infrastructure.

Install OpenSpec CLI (if not already installed) and run openspec init --tools claude to add spec-driven development skills to this project.

Scaffold or show the .claude/project-context.json file that the engineering hooks read to inject project context, target-repo guidance, and optional post-edit format commands.

Scaffold the rules-ex extended-rules infrastructure (.claude/rules/rules-ex-authoring.md and .claude/rules-ex/README.md) required by the engineering plugin's inject-extended-rules hook.

Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.

Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices.

Turn the current conversation into a PRD and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.

Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.

簡潔なコミットメッセージを生成する。Conventional Commits、件名50文字以内、「何を」より「なぜ」。 「コミットメッセージ作って」「/persona-commit」で起動。

自然言語のメモリファイル（CLAUDE.md、todo、設定メモ）を圧縮して入力トークンを削減する。 コード・URL・パス・構造は完全保持。「/persona-compress <path>」「メモリファイル圧縮」で起動。

圧縮出力の subagent への委譲判断ガイド。investigator（位置特定）、builder（1-2ファイル編集）、 reviewer（diff レビュー）。「コンテキスト節約」「圧縮 agent」で起動。

persona プラグインの全コマンド・レベル・キャラクター追加手順のリファレンスカード。 1回限りの表示で状態は変えない。「/persona-help」で起動。

新しいペルソナのキャラクターを作成する。テンプレートから character.md を起こし、 ユーザー層に保存する。「キャラクターを追加」「/persona-new」で起動。

簡潔な PR レビューコメントを生成する。1行1指摘、位置・問題・修正、重大度タグ付き。 「PRレビューして」「コードレビュー」「/persona-review」で起動。

現セッションの実トークン使用量と推定削減量を表示する。セッションログから直接読み取る。 「/persona-stats」で起動。