apollo-toolkit

Reads SPEC.md, validates technical feasibility through web research, finds quality reference implementations, confirms tech stack compatibility, then produces DESIGN.md (architecture design, external dependencies), CHECKLIST.md (verification strategy), Architecture Diff, and design-time refactoring plan. Not for use without a SPEC.md.

Helps users clarify vague requirements through structured conversation, producing a high-level design (PROPOSAL.md) with requirement summaries. Does not read any repo files — relies entirely on conversation. Suitable for requirement discussions, feature ideation, or greenfield project planning.

Executes FIX.md (from the qa skill) as the fix coordinator. The coordinator reads the fix plan, dispatches fix workers and regression test workers, verifies batch gates, and manages the execution flow — it does not write or fix code itself.

Executes PROMPT.md (from the plan skill) as the implementation coordinator. The coordinator reads the plan, dispatches workers, verifies batch gates, and manages the execution flow — it does not write implementation code itself.

Converts SPEC.md + DESIGN.md + CHECKLIST.md into a self-contained coordinator prompt (PROMPT.md) with complete task decomposition, dependency analysis, batch scheduling, and pre-written worker prompts. The generated PROMPT.md is consumed directly by the implement skill.

Reads spec documents and the review output REPORT.md, then generates a self-contained fix coordinator prompt (FIX.md) with issue inventory, dependency analysis, batch scheduling, regression test design, and pre-written worker prompts. The generated FIX.md is consumed directly by the fix skill.

Review a pull request — interactive PR selection via `gh`, 4-dimension code review (hallucinated code, architecture, performance, test validity), then post severity-graded comments with fix suggestions on the PR. Not for spec-based review — use `review` instead.

Reviews spec-related code changes across six dimensions and produces a structured review report (REPORT.md) with findings only — no fix suggestions. Not for non-spec changes, direct code modification, or contexts without a spec.

將已完成的spec歸檔到 `docs/archive/` 下。當你需要將spec歸檔時，調用這個技能。

提交指引與提交前的必要品控閘門。將變更提交到 git repo 或推送到 remote 前，需完成所有品控項目。包括審查、文檔同步與 changelog 門檻。

以 repo 程式碼為唯一依據，維護標準化的 docs/features/、docs/architecture/、docs/principles/ 文檔， 並清理已失效或重複的舊文檔。每條文檔記述必須可追溯至實際程式碼證據。

Initialize the project architecture atlas. Use the apltk CLI to map feature and submodule relationships into a renderable HTML architecture diagram following the C4 model (Context → Container → Component → Code).

Reads the latest codebase and docs to update CLAUDE.md and AGENTS.md at both the project root and per sub-module (crate, package, etc.). CLAUDE.md targets Claude Code; AGENTS.md targets non-Claude agents (Codex, etc.).

Analyzes SKILL.md and supporting files, then produces an optimized rewrite that is clearer, more concise, and gives the agent more freedom to adapt.

Read and search remote GitHub issues via GitHub CLI (`gh`). Use when users ask to list issues, filter issue candidates, inspect a specific issue with comments, or gather issue context before planning follow-up work. Prefer the bundled CLI tools when they are present and working, but fall back to direct `gh issue list` / `gh issue view` commands when the tools are missing or fail for repository-specific reasons.

Transforms user requirements into strictly-scoped business specification documents (SPEC.md). Deeply explores the codebase to calibrate every requirement against actual code. Produces batch specs when requirements exceed 5 BDD items. Not for discussion without PROPOSAL.md, nor for single-file changes that don't need a spec.

Systematically investigates unexpected behavior — reproduces the issue by writing a test, fixes the root cause, and establishes a regression test. Not for simple fixes where the root cause is already known, nor for intuitive fixes without test verification.

Designs test strategy based on risk analysis — selects the right test level and method for the current context. Not for exploratory tasks without tests, nor for tasks where the test type is already predetermined.

Incrementally refresh the architecture atlas when the project diagram drifts from actual code. Measures drift before updating to determine scope, then updates the base atlas and re-renders HTML.

協助完成自動化版本發佈。同步文檔、更新版本號、推送 tag 並建立 GitHub Release。