v2.1.8

Added

• add openclaw as a first-class install target
• add bin/adapters/openclaw.js for OpenClaw CLI/config detection and workspace resolution
• add OpenClaw smoke coverage for default install, custom agents.defaults.workspace, and uninstall restore

Changed

• install shared skills into ~/.openclaw/skills/
• write OpenClaw runtime rules to workspace AGENTS.md
• write persona + output style to workspace SOUL.md
• update README / docs / CLAUDE / pack bootstrap snippets for OpenClaw install and uninstall flow
• extend built-in output styles to support openclaw

Fixed

• make pack bootstrap snippet generation fall back safely when older lock structures do not contain the new host yet

Verification

• npm test -- --runInBand
• npm run verify:skills
• smoke validated for OpenClaw install / custom workspace / uninstall restore

v2.1.7

[2.1.7] - 2026-04-29

Fixed

• 修正 GitHub Actions smoke workflow 断言，使 Claude / Gemini 安装验证与“core skills 默认不暴露 commands”的当前行为一致。
• CI 安全扫描改为在 workflow 中排除 skills/domains/office/，避免新并入的 Office 上游工具链脚本导致 core release 门禁误报。
• 将上述 CI 修复纳入正式发布，避免 v2.1.6 发布后仓库代码与 Actions 结果短暂失配。

Verification

• GitHub Actions: main CI 全绿（含 test matrix + Claude/Codex/Gemini smoke）
• Jest: 22 suites / 220 tests passed（1 skipped）
• Skill contract gate: npm run verify:skills — 26 skills 通过

v2.1.6

v2.1.6

v2.1.4

What's New in v2.1.4

Improvements from PR #17

• skill-registry: new kind inference for routers, workflows, guards directories
• skill-registry: permissions field fallback for allowed-tools, camelCase argumentHint fallback
• utils: rmSafe retry with clearReadOnlyRecursive for Windows compatibility
• install: getSkillPath path normalization for Windows path.sep
• manifest: gemini host files declaration added to packs/abyss/manifest.json
• gemini adapter: reads core files from manifest via getPackHostFiles() instead of hardcoding

SKILL.md Optimization

• All SKILL.md files slimmed down (total 1482 → 851 lines, -43%)
• Detailed content extracted to references/ support files
• Follows official recommendation of keeping SKILL.md under 500 lines

Security Packaging

• Security skill descriptions reframed from offensive to defensive engineering perspective
• Every attack technique now includes detection indicators and mitigation recommendations
• Payloads sanitized with RFC 5737 reserved addresses and placeholder credentials
• Context-appropriate labels (Security Testing, Defense Validation, Security Research)

Codex Install/Uninstall Fix

• AGENTS.md now correctly tracked in install manifest
• Uninstall properly handles {root, path} manifest entries across managed roots
• Empty parent directories pruned after cross-root entry removal
• CI smoke assertions corrected to match actual install layout

Documentation

• README fully rewritten with bilingual (English + Chinese) content
• Complete persona and output style tables
• Accurate install layout diagrams for all 3 targets

CI

• All 12 CI jobs passing (Node 18/20/22 + smoke tests on Linux/macOS/Windows)

---

v2.1.4 更新内容

PR #17 改进提取

• skill-registry：新增 routers、workflows、guards 目录类型推断
• skill-registry：permissions 字段作为 allowed-tools 的回退，支持 camelCase argumentHint
• utils：rmSafe 增加重试机制和 clearReadOnlyRecursive，修复 Windows 兼容性
• install：getSkillPath 路径标准化，兼容 Windows path.sep
• manifest：packs/abyss/manifest.json 补全 gemini host 文件声明
• gemini adapter：从 manifest 读取核心文件映射，消除硬编码

SKILL.md 瘦身

• 所有 SKILL.md 文件精简（总行数 1482 → 851，减少 43%）
• 详细内容拆分到 references/ 支持文件
• 遵循官方建议，SKILL.md 控制在 500 行以内

安全技能包装

• 安全技能描述从攻击视角重构为防御工程视角
• 每项攻击技术均附带检测指标和缓解建议
• Payload 使用 RFC 5737 保留地址和占位符凭证脱敏
• 使用场景标注（Security Testing、Defense Validation、Security Research）

Codex 安装/卸载修复

• AGENTS.md 正确记录到安装 manifest
• 卸载正确处理跨 managed root 的 {root, path} 条目
• 删除跨 root 条目后自动清理空父目录
• CI smoke 断言修正为匹配实际安装布局

文档

• README 全面重写，中英双语
• 完整的人格和输出风格对照表
• 三端准确的安装布局图

CI

• 12 个 CI 任务全部通过（Node 18/20/22 + Linux/macOS/Windows 三平台 smoke 测试）

v2.0.8

v2.0.8

Added

• 新增 target-registry，集中维护 Claude / Codex / Gemini 的 target 与安装根，安装器、style registry、pack registry 与 bootstrap 文档共用同一真相源。
• 新增 config/instruction.md，Codex 安装时会同步到 ~/.codex/instruction.md。
• 新增 Windows CRLF 回归测试，覆盖 gstack frontmatter 解析链路。

Changed

• README 补齐 v2.0.8 新特性、Gemini host 叙事、Codex instruction 同步说明与 npm 安装提示。
• runtime guidance 进一步瘦身，并通过测试门禁限制各风格 guidance 长度预算。
• 历史 automation 分支内容已收束并回到单主线发布。

Fixed

• 修复 Windows 下 gstack frontmatter 只识别 LF 导致的 Claude / Codex / Gemini smoke 失败。
• 修复 pack registry 测试未覆盖 Codex instruction.md 映射导致的 CI 红灯。
• 修复 Codex 默认配置未自动写入 model_instructions_file = "./instruction.md" 的问题。

Verification

• 本地 Codex / Claude Code / Gemini CLI 三端安装验证通过。
• GitHub Actions 全矩阵 CI 通过。
• npm 已发布：code-abyss@2.0.8

v2.0.2

Code Abyss v2.0.2

v2.0.2 聚焦“多风格切换”和“安装链去硬编码”，让 Claude / Codex 两端都能在安装时选择输出风格，并把 Codex 的 AGENTS 生成从静态快照切换到动态拼装。

Highlights

• 新增输出风格 registry
• 支持 --list-styles
• 支持 --style <slug>
• Claude 安装时写入所选 outputStyle
• Codex 安装时动态生成 AGENTS.md
• 新增 3 个额外风格：abyss-concise、abyss-command、abyss-ritual

Added

• bin/lib/style-registry.js
• output-styles/index.json
• output-styles/abyss-concise.md
• output-styles/abyss-command.md
• output-styles/abyss-ritual.md
• test/style-registry.test.js

Changed

• bin/install.js 新增 style 选择参数与风格列举
• Claude 不再把 outputStyle 写死为 abyss-cultivator
• Codex 不再直接拷贝静态 config/AGENTS.md
• README / DESIGN / smoke tests 已同步到多风格模型

Fixed

• 消除“Claude 可切风格、Codex 固定风格”这种双端不对称风险
• 消除安装链中对 abyss-cultivator slug 的单点硬编码，降低后续扩展新风格时的改动面

Verification

• Jest: 13 suites passed, 147 tests passed
• npm run verify:skills
• node bin/install.js --list-styles
• node skills/tools/verify-quality/scripts/quality_checker.js .

Upgrade Notes

• 默认风格仍为 abyss-cultivator
• 不传 --style 时保持兼容
• 可通过 npx code-abyss --list-styles 查看当前可用风格