Development

Use when C or C++ code needs memory-safety or undefined-behavior guarantees proved with CBMC, or ACSL contracts checked with Frama-C Eva or WP. Not for choosing the proof policy: use proof-driven.

Configure optional customer-owned Application Insights telemetry for a Power Apps mobile app — enable it against an existing Azure resource, change the resource, or disable it. Standalone, and invoked by /edit-app.

Generate a self-contained JavaScript server runtime and registration metadata for an MCP codeful tool. Use when the user asks to create a codeful MCP tool, generate server logic for an MCP tool, write a runTool function, build a Dataverse-backed MCP tool, or pair MCP server logic with an MCP App widget.

Production-grade Android app development guide covering native (Kotlin/Java), cross-platform (Flutter, RN, KMM), and hybrid architectures.

Use when Josh asks to write a note, make a note, or capture something for his "pickled-knowledge" Obsidian vault - triggers on "write a note about X", "make a note for this", "capture this", "note this up", "put together a note on X", or a bare URL/article/tweet handed over with intent to write it up. Do NOT use for generic writing help, drafting prose that isn't a vault note, or when Josh is just asking a question and not asking for a note.

仕事の日本語文書を読みやすくわかりやすく書く・直すためのスキル。議事録（文字起こしからの議事録化を含む）、調査レポート・分析レポート、社内ガイド・マニュアル、リサーチメモ・ディスカッションペーパー・企画書・提案書・報告書・メール、スライド構成案といったビジネス文書の作成・校正、「結論から書いて」「論旨を明確に」「見出しを端的に」「専門用語をわかりやすく説明して」といった指示のいずれでも使用する。AI臭さの除去（「AIっぽい」「AI臭い」「機械翻訳っぽい」「不自然」「もっと自然な日本語に」「機械っぽい」「人間っぽくして」「単調」「〜することができる、と言えるだろう、のような言い回し」といった直接・間接・口語の指摘、AIで書いたと言われた/疑われた）、読みにくい・わかりにくい文章の改善依頼（語順がおかしい、一文が長い、何が言いたいか分からない、読点の位置がおかしい等）、note記事やブログ記事・エッセイの新規執筆（任意のテーマをゼロから書く・書き起こす依頼を含む）、既存文章のリライト・推敲、自分の文体を学ばせたい・プロファイル化したいという要望（過去の文章を読ませて自分らしく書いてほ...

Bring a workhub vault's layout up to what the current app and plugins expect, after a structural change. Use when hooks stop injecting owner context, when the app reports a breaking change, or when a vault has not been opened in a long time.

Record and use a team knowledge base that lives outside the vault - a network drive, Google Drive, OneDrive or SharePoint folder. Survey mode reads the place and writes down how it is organised into projects/NNNN-<slug>/shared/; place mode says where a vault note belongs in it. Use when the user wants to register a team share for a project, refresh what is recorded about one, or work out where to file something.

Convert an HTML file into a publication-quality PDF with headless Chromium — a document paginated by @page CSS with margins, bookmarks and page numbers, or a fixed-stage slide deck captured one slide per page. Use when the user wants to export, print or hand over an HTML report, one-pager or presentation as a PDF, or asks to turn a web page into a PDF.

Build a self-contained HTML document people actually read — an explainer for a customer or a colleague, a one-pager, a status or research report, a dashboard, a comparison — with a real design system behind it and a PDF to hand over. Use when a Markdown document would be too long or too flat for its audience, when answering an enquiry with something readable, or when the user asks for a report, a summary or a one-pager.

Draw a diagram as a self-contained HTML file with inline SVG, in an editorial style with orthogonal connectors and a hard complexity budget — for explaining a question or an answer to someone, working out a system, or organising an idea. Use when the user wants a flowchart, sequence, architecture, timeline, tree, swimlane, quadrant or cycle diagram, wants something explained visually, or is thinking a plan through and needs it laid out.

Build a presentation as one self-contained, animation-rich HTML file on a fixed 16:9 stage, and export it to PDF — the whole deck rendered here, with no external design tool in the loop. Use when the user wants slides, a deck, a talk, a pitch or a presentation, wants an existing HTML deck reworked, or asks how to hand slides over as a PDF.

Quiz the user on a document, one question at a time, until they can explain it to someone else — definitions, numbers, evidence, background, scope — and log the document's own gaps separately. Use when the user wants to understand material rather than have it summarised, is preparing to explain or present a document, or says 「理解できるまで質問して」「説明できるようになりたい」「この資料を叩き込みたい」「teach-back」.

Structure the user's thinking one agreed level at a time, by asking rather than answering. `down` breaks an abstract concept ("strategy", "AI adoption") into definition, purpose, desired state, components and observable conditions; `up` climbs from a symptom on the ground (reviews stall, sprints overrun) through facts and causes to the purpose and the state worth aiming for. Use when the user wants to think a concept or a problem through themselves, or says 「一緒に整理して」「分解したい」「本質から考えたい」「なぜ起きているの...

Show what is going on in the team's comms space - live threads, their state, and what is unread for you - and help pick one to work in. Use when the user asks what the team is discussing, what they have missed, or which thread to join.

Put this session's research, comparison, or investigation into a team-comms thread as a summary plus attachments, so teammates' agents can read it instead of being handed a document. Use when work worth sharing with the team comes out of a session, or when the user would otherwise export a document and send it.

Write a reply or an independent opinion into a team-comms thread. Use when the user wants to answer a question from another member's agent, add a view to a discussion, or respond to something surfaced as unread.

Start a new discussion thread in the team's comms space so several members' agents can weigh in. Use when the user wants to ask the team something, raise a design question, or get other agents' opinions on a decision.

Declare that this working directory is now taking part in a team-comms thread, so future sessions here surface its unread posts - or stop, so they go quiet again. Use when the user says they are working in a thread, wants notifications for one, or wants them to stop.

Record the conclusion of a team-comms discussion and close the thread. Use when a discussion has reached an answer, when the user says what was decided, or when a thread is finished and should stop appearing as live.

Restore a team-comms thread's discussion into this session - the cached digest plus whatever arrived since - and save an updated digest. Use when starting work in a thread whose history this session does not have, after focusing one, or when the user asks what has been said.

Connect this machine and project to a team-comms discussion space on a shared folder - create .claude/team-comms.json, verify the drive is really readable, and register this agent. Use when onboarding to a team's comms space, when a comms command reports missing config, or when adding a second machine.

Keep memory small enough to stay useful - report which caps are over, propose merges, split bloated notes, archive what is finished, and connect orphans. Use when memory-doctor or the session brief reports a cap, when memory feels noisy, or on a periodic review.

Promote what a session learned from the short-term record into durable memory - review recent episodes and conversations, keep what will matter next time, and write it as typed notes. Use at the end of a session, when the user asks to reflect/consolidate/remember what was learned, or when memory-doctor reports the episodes cap.