Turn any idea, plan, or codebase into a beautiful interactive diagram. An agent skill for Claude Code, Codex, and more.
HTML agent-skillsai-agentsarchitecture-diagramclaude-codeclaude-skillscodexcoding-agentsdeepseek-harnessdeveloper-toolsdiagramsdiagrams-as-codedsh-pluginflowchartllmmermaidopencodesequence-diagramsoftware-architecturesystem-designvisualization
Merge branch 'dev' into feat/283-hermes-agent-install
Sayujaya14676e17536443
diff --cc CHANGELOG.mdindex e8cf9ca7,eb7c3bc1..cc53f4a6--- a/CHANGELOG.md+++ b/CHANGELOG.md@@@ -7,9 -7,11 +7,12 @@@ All notable changes are documented here> Development identity: `v2.17.0-dev.1`. Not a stable release.### Added+- **Hermes Agent community opt-in.** `integrations/hermes-agent` is a Skill-only directory plugin: it registers the existing Node Archify `SKILL.md` for Hermes. The documented install is `hermes skills install skills-sh/tt-a1i/archify/archify -y` (published GitHub Skill). A checkout symlink remains the local-dev path. Hermes still runs `node bin/archify.mjs`. This is not an official Nous product and is not an agent-switcher target.+ - **GitHub Copilot agent-switcher target.** The Start page agent switcher, README EN/ZH quick start, and landing copy add `github-copilot` alongside `cursor`, `codex`, `claude-code`, and `opencode`. It installs the same checked Skill and zero-dependency renderers via `npx skills add tt-a1i/archify --skill archify --agent github-copilot ...`; no vendor-specific fork, renderer, or schema behavior was added.### Fixed+ - **Conservative VS15 text-width estimates.** Text-presentation selectors retain the base character's estimated width instead of forcing every sequence to one unit. This prevents wide labels, including CJK text whose font ignores VS15, from passing layout checks at half their estimated width. Actual glyph widths remain font-dependent; some text-presentation labels may now need more space. VS16 emoji estimates and text without selectors are unchanged.+ - **Readable renderer failures.** `render` and direct renderer entry points format classified input, schema, layout, and output failures without Node stacks. Filesystem failures identify the failing operation instead of labelling output errors as unreadable input; successful artifact bytes and existing validation/delivery receipt formats remain unchanged.- **Architecture Delta baseline arrowheads (#433).** Removed and rerouted baseline relationships retain their marker definitions in the composed Delta SVG, preserving their authored direction alongside current relationships.- **Compare rollback recovery (#438).** If restoring a previous output fails, compare preserves its recovery directory and reports backup-to-target paths instead of deleting the remaining backups during cleanup.- **固定提交的来源校验 (#420)。** 校验忽略本地 Git replacement refs,始终读取指定提交的原始对象,避免替换后的文件或行范围造成误接受或误拒绝;保留原有来源链接、诊断与用户 Git 配置。diff --cc README_ZH.mdindex c19e4380,7b462b34..34177501--- a/README_ZH.md+++ b/README_ZH.md@@@ -109,7 -109,7 +109,8 @@@ npx -y skills add tt-a1i/archify --skilnpx skills use tt-a1i/archify@archify --agent codex```+DeepSeek Harness(社区集成、显式启用):运行 `dsh plugin --profile web add @tt-a1i/[email protected]`;参见[兼容范围、限制与安全说明](integrations/deepseek-harness/README.md)。Hermes:`hermes skills install skills-sh/tt-a1i/archify/archify -y`;参见[说明](integrations/hermes-agent/README.md)。[Agent 切换器](https://tt-a1i.github.io/archify/start.html?agent=cursor&type=architecture)只为 `cursor`、`codex`、`claude-code` 和 `opencode` 生成命令。Raven 仅支持 ZIP 手动安装:将 [`archify.zip`](archify.zip) 解压到 `~/.raven/workspace/skills`,解压后会得到 `~/.raven/workspace/skills/archify`;Raven 不属于切换器目标。+ DeepSeek Harness(社区集成、显式启用):运行 `dsh plugin --profile web add @tt-a1i/[email protected]`;参见[兼容范围、限制与安全说明](integrations/deepseek-harness/README.md)。[Agent 切换器](https://tt-a1i.github.io/archify/start.html?agent=cursor&type=architecture)只为 `cursor`、`codex`、`claude-code`、`opencode` 和 `github-copilot` 生成命令。Raven 仅支持 ZIP 手动安装:将 [`archify.zip`](archify.zip) 解压到 `~/.raven/workspace/skills`,解压后会得到 `~/.raven/workspace/skills/archify`;Raven 不属于切换器目标。安装后的 Skill 包含一个低频、失败静默的发布检查,它最多只显示可选更新提醒,绝不会自行下载或安装更新。一次成功检查后,下次网络请求通常约在 72 小时(±20%)后发出;检查失败后,活跃使用可能在首次 6 小时、后续 24 小时退避到期时重试。请求只访问 `https://tt-a1i.github.io/archify/skill-updates/archify/stable.json`。服务端会自然获得 IP、请求时间和常规 HTTP 元数据;检查器不会发送本地版本、Agent、项目数据、用户输入、账户/设备标识,也不会保存或回传 ETag。是否更新以及何时更新始终由你决定。如需完全关闭检查(包括网络请求和提醒状态写入),请在 Agent 环境中设置 `ARCHIFY_UPDATE_CHECK_DISABLED=1`。