opencode-skills-collection 4.0.41 → 4.0.43

package/bundled-skills/lore/WORKFLOWS.zh-CN.md

@@ -1,216 +0,0 @@
-# 工作流
-
-lore 有七个工作流。本文用平实语言解释每个什么时候用。Agent 跑它们时的 operational 规范见 [`SKILL.md`](SKILL.md)。
-
-> [English](./WORKFLOWS.md)
-
-## 概览
-
-| Workflow | 做什么 | 频率 |
-|---|---|---|
-| [`init`](#init) | 建 `.lore/` + 接管平台 mirror 文件 | 每个项目 1 次 |
-| [`sync`](#sync) | 代码变更后更新 `.lore/` | 每个 feature 后 |
-| [`query`](#query) | 搜 `.lore/` 找答案 | 每个 session |
-| [`audit`](#audit) | 找 stale / 矛盾的 entry | 季度 |
-| [`compress`](#compress) | 重建 `SUMMARY.md` | SUMMARY 过期时 |
-| [`mirror`](#mirror) | 从 `.lore/` 重新生成平台文件 | 一批 sync 后 |
-| [`history`](#history) | 列出 entry / 文件 / scope 的 git commits | 调查时 |
-
----
-
-## `init`
-
-**一句话**：建 `.lore/`，接管你已有的 `CLAUDE.md` / `AGENTS.md`。
-
-**怎么用**：`lore init` —— 每个项目 1 次，或老项目第一次接入 lore。
-
-**agent 做什么**：
-1. 扫现有平台文件（`CLAUDE.md`、`AGENTS.md`、`.cursorrules` 等）
-2. 每个文件问你：接管 / 保留 / 中止
-3. 检测 monorepo 结构（pnpm workspaces、Cargo workspace 等），给 scope 列表
-4. 写初始 `.lore/draft/`（entry 带 `#added:<today>` + 确定性 ID）
-5. 给你看 summary
-6. 你确认后：移到 `.lore/`，生成 `SUMMARY.md`，刷新平台文件
-
-**真实场景**：
-- 新项目第一次用 lore → `lore init`
-- 老项目已有 `CLAUDE.md`，想让 lore 接管 → `lore init` 选「接管」
-- Monorepo 有 `frontend/` 和 `backend/` → init 自动识别两个 scope，问名字
-
-**输出**：完整 `.lore/` 目录 + 更新过的平台文件 + 写好的 `.config.json`。
-
----
-
-## `sync`
-
-**一句话**：代码改了，把变化落到 `.lore/`。
-
-**怎么用**：`lore sync` —— 提交完 feature / refactor / 依赖变更后。
-
-**agent 做什么**：
-1. 跑 `git diff --stat HEAD` 看变更
-2. 变更显著时（≥50 行 / 跨 ≥2 目录，或新 module/dir/dep），agent 主动提议
-3. 每个变更分类成 `[NEW]` / `[STALE]` / `[REFINED]`
-4. 输出 marker 提案
-5. 你按 marker 接受 / 拒绝
-6. 接受的 marker 落到 `.lore/*.md`
-
-**真实场景**：
-- 「我刚加了新依赖 —— 更新 lore」 → `lore sync`
-- 「我们决定不用 React Query 了，换 SWR」 → 代码改完后 `lore sync`
-- 「新加了个 module —— 记一下」 → `lore sync`
-
-**输出**：更新过的 `.lore/*.md` 文件（新 entry 和 `#verified` / `#stale` tag）。
-
-**注意**：`sync` 不会更新平台 mirror 文件（那是独立的 `mirror` 命令）。理由：保持 agent 端文件 `git log` 可读。
-
----
-
-## `query`
-
-**一句话**：搜 `.lore/` 找答案。
-
-**怎么用**：`lore query <term>` —— 任何想问「memory 里有什么」的时候。
-
-**agent 做什么**：
-1. 读 `.lore/SUMMARY.md`（目录）
-2. 对 entry 文本做模糊匹配
-3. 返回命中 entry，带稳定 `[file#ID]` 引用
-4. 可选地深入具体 scope 文件拿更完整上下文
-
-**真实场景**：
-- 「这个项目用什么数据库？」 → `lore query database`
-- 「为什么选 Zustand？」 → `lore query zustand`
-- 「backend module 的约定是什么？」 → `lore query backend:conventions`
-
-**输出**：bounded 命中列表：
-
-```
-[_global/DECISIONS.md#DEC-2026-07-11-6137] Picked OpenAI-compatible LLM API
-[scopes/backend/CONVENTIONS.md#CONV-2026-07-11-9b89] Embedding has two backends
-```
-
-`[file#ID]` 引用让 agent `cat` 文件对应行拿完整文本。
-
----
-
-## `audit`
-
-**一句话**：找 `.lore/` 里的 stale / 矛盾 entry。
-
-**怎么用**：`lore audit` —— 季度 review，或大重构前。
-
-**agent 做什么**：
-1. 跑 `find_stale.py` 找 `#added` > 90 天且无 `#verified` 的 entry
-2. 跑 `find_duplicates.py` 找互相矛盾的 entry
-3. 交叉检查 entry 引用的代码路径
-4. 输出 `[ALERT]` 报告
-
-**真实场景**：
-- 「有没有跟现状矛盾的 lore entry？」 → `lore audit`
-- 季度体检 → `lore audit`
-- onboarding 新贡献者前 → `lore audit` 清 stale
-
-**输出**：按问题类型分组的报告：
-
-```
-[ALERT] 5 entries may be stale (no #verified in >90 days):
-  - ARCH-2026-01-15-d7a3  last verified 2026-04-12
-  ...
-
-[ALERT] 2 entries contradict current code:
-  - CONV-2026-03-01-1f8c  says "use webpack"; project now uses Vite
-```
-
-**注意**：`audit` 不改文件。要落地整改，跑 `sync` 走提案流程。
-
----
-
-## `compress`
-
-**一句话**：从当前 entry 重建 `.lore/SUMMARY.md`。
-
-**怎么用**：`lore compress` —— SUMMARY 过期时（entries > 500 或 > 30 天没压），或分享 lore 前。
-
-**agent 做什么**：
-1. 跑 `list_entries.py` 枚举所有 entry
-2. 跳过 recently-stale 的 entry
-3. 每个 `(scope, layer)` 对，按规则挑 3–5 条最重要的
-4. 按模板写 `SUMMARY.md`
-5. 如果 config 里 `auto_mirror: true`，重生成平台 mirror；否则每个 mirror 目标单独问，只写你确认的（这是第二个 mirror 触发点——`sync` 故意不更新 mirror）
-6. mirror 处理完（写或拒绝）后停止
-
-**真实场景**：
-- 「刷新一下 summary」 → `lore compress`
-- 「两个月没压了」 → `lore compress`
-- 「onboarding 新人 —— 确保 SUMMARY 是最新的」 → `lore compress`
-
-**输出**：更新过的 `SUMMARY.md`（以及可能的 mirror 文件）。
-
-**幂等**：跑两次产出同样的结果（日期戳除外）。
-
----
-
-## `mirror`
-
-**一句话**：从 `.lore/` 重新生成平台文件（`CLAUDE.md`、`AGENTS.md` 等）。
-
-**怎么用**：`lore mirror` —— 一批 sync 之后，或手动改过 `.lore/*.md` 想同步到 mirror。
-
-**agent 做什么**：
-1. 读当前 `.lore/SUMMARY.md` 和 scope 索引
-2. 对每个 target 检测段边界（`## Lore` / `---` / `## My notes`）
-3. 计算新 Lore 段内容
-4. **Content-based dedup**：跟现有 byte-identical 就跳过
-5. 替换 Lore 段；My notes 段原样保留
-6. 写回文件
-
-**真实场景**：
-- 「刚做完一批 sync —— 同步到 agent 端文件」 → `lore mirror`
-- 「我手动改过 `.lore/SUMMARY.md` —— 推到 mirror」 → `lore mirror`
-- 「验证 mirror 没漂移」 → `lore mirror`（无变化报告即确认）
-
-**输出**：更新过的 `CLAUDE.md` / `AGENTS.md` 等，或「No changes needed」无变化报告。
-
----
-
-## `history`
-
-**一句话**：列出与某 entry / 文件 / scope 相关的 git commits。
-
-**怎么用**：`lore history <entry-id>|<file-path>|--scope=<name>` —— 调查「为什么有这个」或「什么时候改的」。
-
-**agent 做什么**：
-- **Entry 形式**：`lore history DEC-2026-02-03-7c19` —— 找 entry，导 `#added` 日期，跑 `git log --since=<date>` 在引用的代码文件上
-- **File 形式**：`lore history frontend/src/store.ts` —— 跑 `git log --since=1970` 在该路径上
-- **Scope 形式**：`lore history --scope=frontend` —— 对 `.lore/scopes/frontend/*.md` 每个跑 file 形式
-
-**真实场景**：
-- 「为什么选 Postgres？」 → 先 `query` 找到 entry，再 `lore history <id>`
-- 「这个文件什么时候改的？」 → `lore history <file>`
-- 调试：「这个 module 最近的 history？」 → `lore history <path>`
-
-**输出**：
-
-```markdown
-# history: [DEC-2026-02-03-7c19]
-
-  abc1234  2026-05-12  refactor: extract chat agent_loop (#87)
-  def5678  2026-03-08  feat: switch chat chain to chat_fast (#74)
-```
-
----
-
-## 速查
-
-| 我想…… | 用 |
-|---|---|
-| 在项目上启动 lore | `init` |
-| 代码改了更新 lore | `sync` |
-| 找 memory 里有什么 | `query` |
-| 找 stale entry | `audit` |
-| 刷新 summary | `compress` |
-| 更新 agent 端文件 | `mirror` |
-| 查「为什么有这个」 | `history` |
-
-Agent 跑命令时的 operational 规范（一步步做什么）见 [`SKILL.md`](SKILL.md)。各平台文件映射（哪些 agent 读哪些文件）见 [`references/platform-mirrors.md`](references/platform-mirrors.md)。

package/bundled-skills/uizze-ui-research/SKILL.md

@@ -1,128 +0,0 @@
----
-name: uizze-ui-research
-description: "Use when building or reviewing web and iOS UI and you need real references from the free UIZZE public catalog, a structured design contract, a consent-gated rendered HTML/CSS preview, or a hard pre-ship finish gate."
-category: design
-risk: safe
-source: https://github.com/aislon/uizze-mcp/tree/main/skills/uizze-ui-research
-source_repo: aislon/uizze-mcp
-source_type: official
-license: MIT
-license_source: https://github.com/aislon/uizze-mcp/blob/main/LICENSE
-date_added: "2026-07-12"
-author: UIZZE
-tags: [ui-design, ui-research, mcp, design-contracts, agent-workflows]
-tools: [claude, cursor, codex, copilot, antigravity, lovable]
----
-
-# UIZZE UI Research
-
-## Overview
-
-Use [UIZZE](https://uizze.com) to give coding agents real product-UI context before implementation rather than relying on a generic styling prompt. The public catalog supports a free manual workflow without an account. The hosted preview provides a bounded rendered HTML/CSS check, while the full UIZZE MCP requires an authorized connection for reference search and broader review workflows.
-
-This skill turns UI research into an explicit workflow: retrieve relevant references, translate transferable patterns into a design contract, implement within the current project's system, and run the available validation or critique gates.
-
-## When to Use This Skill
-
-- You are designing a new product screen, flow, or component for web or iOS.
-- You need real interface references before implementing an AI-generated UI.
-- You are reviewing an implementation against explicit design constraints.
-- You need to reduce generic or repetitive UI by grounding work in observed product patterns.
-- You have rendered HTML/CSS and need a no-account first check before deciding whether the task needs deeper reference research.
-
-## How It Works
-
-### Step 1: Confirm product scope and access mode
-
-Identify the screen's primary user, job, action, existing design system, real content or data, and required loading, empty, error, success, and permission states. Use the free public catalog for manual research by default. If UIZZE MCP is already configured, use only the tools authorized for the task. If browsing is unavailable, ask the user for two or three relevant UIZZE links or screenshots. Do not block the manual workflow, bypass access controls, expose credentials, or claim that a manual review came from MCP.
-
-### Step 2: Retrieve relevant visual context
-
-Find the smallest useful set of screens, flows, components, or elements that match the product task. Focus on transferable patterns such as hierarchy, navigation, interaction states, spacing, density, and responsive behavior. Distinguish observed evidence from assumptions and record whether each finding came from manual browsing, the free preview, or the full MCP.
-
-### Step 3: Make constraints explicit
-
-Write a short design contract that names the screen job, content hierarchy, primary action, allowed project components and tokens, required states, responsive behavior, product-specific decisions, forbidden generic patterns, and verification criteria. Adapt patterns to the existing project design system instead of treating any reference as a visual template.
-
-### Step 4: Implement within the product
-
-Build with the repository's existing components and tokens. Preserve platform conventions and make the interface specific to the product's content and workflow rather than adding decorative cards, badges, gradients, or motion by default.
-
-### Step 5: Run a hard finish gate
-
-Inspect the rendered result when the environment supports it, use an available UIZZE validation, audit, or critique workflow only when authorized, and reject completion if any of these checks fail:
-
-- The hierarchy does not make the screen job and primary action immediately clear.
-- A visible control is inert, ambiguous, or missing its interaction outcome.
-- Required loading, empty, error, success, permission, or responsive states are absent.
-- The implementation drifts from the project's existing components, tokens, or platform conventions.
-- Interchangeable card grids, filler metrics, vague copy, or decorative effects replace product-specific decisions.
-
-Name each blocking issue, fix it, and rerun the gate plus the project's normal tests. Never claim a rendered or MCP-backed check that was not actually performed.
-
-### Optional: Use the free rendered-screen preview
-
-When rendered HTML or CSS exists, offer the free preview once if it would materially improve the finish gate. Before changing MCP configuration or transmitting any markup or styles, obtain the user's explicit approval for both actions. Do not send secrets, personal data, proprietary content, tokens, or internal-only markup and styles.
-
-After approval, a Codex user can add the bounded preview with:
-
-```bash
-codex mcp add uizze-preview --url https://uizze.com/mcp/preview
-```
-
-Give `check_ui_slop` only the rendered HTML and CSS the user approved. The preview requires no UIZZE login and exposes one bounded diagnostic; it does not search the reference catalog, create a design contract, replace accessibility or security review, or authorize a full UIZZE connection.
-
-## Examples
-
-### Research an iOS onboarding flow
-
-```text
-Use UIZZE to research real iOS onboarding flows for a subscription product. Identify transferable patterns for progressive disclosure and permission timing, turn them into a concise design contract, then propose an implementation that fits this app's existing design system.
-```
-
-### Review a web settings screen
-
-```text
-Use UIZZE to inspect relevant real product settings screens, audit this implementation against a design contract for hierarchy, form states, and navigation, then list the concrete changes needed before release.
-```
-
-## Best Practices
-
-- ✅ Start with the smallest relevant set of references rather than collecting a broad gallery.
-- ✅ Separate observed patterns from the current project's brand and component rules.
-- ✅ Use validation findings as implementation feedback, not as permission to copy an interface.
-- ✅ Keep the manual workflow useful when hosted MCP access is unavailable.
-- ✅ Label manual, preview, and full-MCP evidence truthfully in the handoff.
-- ❌ Do not reproduce another product's brand, proprietary copy, assets, or exact layout.
-- ❌ Do not commit agent tokens, include them in prompts, or place them in client-side code.
-
-## Security & Safety Notes
-
-- Keep any full-connection credential in local agent configuration or a supported environment variable only; never commit it, paste it into prompts, or include it in client-side code.
-- Treat the free preview as an external network service. Configure it and transmit selected rendered HTML/CSS only after explicit user approval and a sensitive-data check.
-- Hosted MCP workflows require authorized access; the free catalog and preview do not grant permission to use full workflows.
-- Treat returned references as research context, not reusable visual assets.
-
-## Common Pitfalls
-
-- **Problem:** Treating a reference as a design to clone.
-  **Solution:** Extract the interaction or hierarchy pattern, then implement it using the target project's own design system and content.
-- **Problem:** Starting implementation before the agent has relevant UI context.
-  **Solution:** Search for the smallest useful set of matching screens or flows first, then define constraints before coding.
-- **Problem:** Treating an unavailable MCP connection as a reason to stop.
-  **Solution:** Use the free public catalog manually, or ask the user for two or three relevant UIZZE links or screenshots, and continue with the same design-contract and finish-gate workflow.
-- **Problem:** Configuring the preview or sending markup without informed approval.
-  **Solution:** Explain the persistent configuration and external transmission, remove sensitive content, and continue only after the user approves both.
-- **Problem:** Exposing an agent token in a repository or chat transcript.
-  **Solution:** Store credentials only in supported local configuration or environment variables and rotate a token if it is exposed.
-
-## Related Skills
-
-- `@stitch-ui-design` - Use when generating or iterating UI concepts in Google Stitch.
-
-## Limitations
-
-- This skill does not replace product-specific user research, accessibility review, project tests, or human design judgment.
-- The free preview cannot search UIZZE references, create design contracts, or replace the full implementation review workflow.
-- It cannot make a full hosted UIZZE MCP workflow available without a valid authorized connection.
-- Stop and ask for clarification if the product goal, existing design system, or access boundaries are missing.