fstack

Document codebase as-is via parallel sub-agents. Use when researching how code works, mapping architecture, or answering questions about existing systems. Auto-detects thoughts/ directory — if present, integrates historical context and syncs; if absent, skips those steps.

Create the Questions artifact before codebase research, design, planning, or implementation. Use for feature tickets, architecture changes, unclear code work, CRISPY/QRSPI Questions phase, or when the agent needs to define code-answerable unknowns and human questions before researching or planning.

Create a concise Design Concept artifact after questions and codebase research are available. Use for CRISPY/QRSPI Design phase, architecture alignment, feature design discussion, or turning current-state research into a small human-reviewable proposed shape before detailed planning.

Create an interface-first Structure Outline before detailed planning or implementation. Use for CRISPY/QRSPI Structure Outline phase, vertical planning, method signatures, type contracts, API/component boundaries, or converting an accepted design concept into end-to-end slices with verification checkpoints.

Create comprehensive handoff documents for transferring work to another agent session. Use when the user requests to create a handoff, end a session, or document work-in-progress for future continuation.

Create detailed implementation plans through interactive research and iteration. Use when starting a new feature or ticket that needs a comprehensive technical plan. Saves plans to the repo-appropriate plans location and emits Personal Knowledge events when applicable.

個人腦力激盪教練。用結構化技巧逼出你自己的想法，而非替你產生點子。 僅在使用者明確說出「腦力激盪」或「brainstorm」時觸發，其他模糊說法（幫我想想、 想不出辦法、有什麼可能）不觸發。 角色：Claude 是 coach，用 probing question 引導，不是 idea generator。

通用 skill 自我進化工具。當任何 skill 執行後使用者有修正、抱怨、或說「少了什麼」， 且能確認是可重複的 skill 規則缺口時，才在 response 末尾輸出 <<GAP skill-name: 一句話描述缺口>>， Stop hook 自動寫入 signal-queue.md。 主動觸發：「improve」「skill 改進」「improve skill」「修正 skill」「更新 skill 規則」 「skill 有缺口」「skill 漏掉了」「self-improve」「自我進化」「skill 學習」。 初始化：「improve init」批量為選定 skill 生成 Signal Collection 區塊（供人工審核）。

Mine mechanism-first candidate claims from a bounded source set across thoughts/ and Heptabase MCP. Writes candidate files to thoughts/global/fredrick/candidates/ and stops for human scoring. Trigger: "personal-wiki-mine", "zettel mine", "candidate claims", "卡片盒連結挖掘".

Reflect on /improve Skill Evolution Memory. Use when the user asks to process accumulated improve signals, deduplicate skill gaps, generate or refresh memory/claims/{skill}.md, generate memory/eval-cases/{skill}.json, enrich old memory records, or produce a prioritized /improve worklist from signals.jsonl.

每日早晨任務看板：整合 gsheet / GitHub / Heptabase Sprint whiteboard + week tag 三處任務狀態，自動建卡、合併狀態、寫入 Heptabase 當日 journal。 觸發詞：「start-day」「每日任務看板」「今日任務」「我要開始工作」 「daily task board」「整理今日任務」。

在 Heptabase 建立符合 Sprint whiteboard 標準格式的任務卡片。 當使用者說「建立任務卡片」「新增 sprint card」「create task card」「加一張卡到 sprint」， 或要把計畫拆解成子任務時立即觸發。 支援批次建立多張卡片，並在每張卡片的 `## Depends on` section 標記依賴關係， 讓 /start-day skill 可正確解析。

Applies the Deep Module principle (John Ousterhout / Matt Pocock) at any stage of development. Use this skill during planning (create-plan) to shape module boundaries before writing specs, during implementation (implement-plan) to enforce interface-first design before writing code, or during code review to identify shallow abstractions. Trigger when the user mentions "deep module", "shallow module", "interface design", "encapsulation", asks to review module structure, or when any new module/class/function is about to be designed or implemented.

把模糊的功能想法逼成具體的 PRD。一次問完所有關鍵問題，存成 md 檔讓使用者 填答，迭代到對齊後產出 PRD draft，再交給 create-plan 寫詳細計畫。當使用者 說「grill me」、「拷問」、「釐清需求」、「想做 X 但很模糊」、「幫我把想法 變具體」、「pre-PRD」時觸發。**只做需求對齊，不寫程式、不寫 plan**。

Implement an approved technical plan from thoughts/shared/plans/. Use when executing a plan that was created with create-plan. Follows phases, updates checkboxes, and pauses for human verification between phases. Reports a structured walkthrough summary inline at the end.

Read a standard plan and decompose it into a team-plan with task briefs, dependency graph, and model assignments. Use when you want to run a complex plan with agent teams. Outputs a team-plan file for /implement-team-plan to execute. Does NOT spawn teammates — that's implement-team-plan's job.

Research the codebase and create a concise implementation plan in one step. Use for small, focused changes where you want to go straight from research to a ready-to-implement plan without back-and-forth. For larger or more complex changes, use /research-codebase followed by /create-plan instead.

Resume work from a handoff document with context analysis and validation. Use when the user requests to resume from a handoff, continue previous work, or references a handoff path or ticket number (ENG-XXXX). Supports quick mode (default) and full mode (--full flag).

Render beautiful, professionally-styled Mermaid diagrams using the beautiful-mermaid npm package. Use when the user asks to "create a beautiful diagram", "create a styled diagram", "render a Mermaid diagram", or requests visual diagrams with theming. Supports flowcharts, state diagrams, sequence diagrams, class diagrams, and ER diagrams with 15 built-in themes.

Adversarial brainstorming. Claude and Codex independently research then debate until Nash equilibrium. For solution exploration, feasibility analysis, exhaustive enumeration.

Use Codex CLI (not MCP) to review uncommitted changes. Codex explores the codebase independently with full disk read access.

從 candidate-queue.md 讀取待審候選 skill，進行去重、合併、review， 使用者批准後呼叫 skill-creator 實際建立。 觸發詞：「extract-skill」「抽取 skill」「建立 skill 從 queue」「處理 skill 候選」 「有 skill 候選待建立」「extract skills from queue」「把候選做成 skill」

Execute a team-plan (produced by /create-team-plan) using Claude Code's native agent teams. Spawns teammates per task brief, manages execution, and produces a walkthrough artifact. Requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 and Claude Code v2.1.32+. Use after /create-team-plan has produced a team-plan with status: ready.

功能完成後的收尾三連發：commit+PR → gsheet 進度同步 → (選配) Slack 通知。 觸發詞：「work-wrap-up」、「收尾」、「功能做完了」、「幫我收尾」、「wrap up」、 「wrap」、「做完了」、「push 完了」、「收工」。 依序執行：commit-commands:commit-push-pr → gsheet-progress-sync → (選配) slack-message。