CCDawn/briefbound-skills / skills/engineering/briefbound-router
briefbound-router
VerifiedUse when Chinese-first active intent discovery, local context probing, user calibration before downstream skill selection, owner/workflow routing, continuation, permission inference, or multi-thread/conflict coordination is needed.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.8 KB |
| agents/openai.yaml | config | 375 B |
| references/capability-routing.md | markdown | 5.2 KB |
| references/collaboration-discovery.md | markdown | 1.7 KB |
| references/evaluation.md | markdown | 989 B |
| references/github-skill-candidates.md | markdown | 11.0 KB |
| references/harness-compat.md | markdown | 3.6 KB |
| references/output-forms.md | markdown | 2.9 KB |
| references/review-test.md | markdown | 2.6 KB |
| references/routing-practice.md | markdown | 9.8 KB |
| references/runtime.md | markdown | 3.2 KB |
| references/ui-preview-approval.md | markdown | 3.3 KB |
references/harness-compat.md
View on GitHub# Harness 能力矩阵与降级
Briefbound 技能包运行在多种 agent 运行环境(Codex App、Codex CLI、ZCode、Claude Code、Cursor、Grok Build、Gemini CLI、OpenCode 等)。加载协作族技能或写入隔离规则前,先按本文件探测当前环境能力;缺失能力按表降级,不模拟协议、不发明工具名。
## 能力分级
- `C1 基础执行`:文件读写、命令执行、Git。所有 harness 默认具备;本包全部技能以 C1 为底线。
- `C2 skill 发现`:SKILL.md 目录(agentskills.io 开放格式)。由安装器装入各 harness 的 skills 目录后具备。
- `C3 子代理派发`:runtime 提供任务委派工具(如 ZCode Agent、Claude Code subagent)。只决定独立可并行工作能否派发;与平级会话协作无关,子代理不是平级会话。
- `C4 平级多会话`:同项目独立会话枚举与定向消息(如 Codex App 的 `list_threads`/`read_thread`/`send_message_to_thread`)。BRT relay 协议族(`BRT_TRUSTED_RELAY_V1`、outbox、`resumePendingAgentIds`)只在 C4 成立时生效。
- `C5 协调 registry`:`agent_coordination.py` + `briefbound/coordination/registry.json`(claim、租约、恢复债务)。需要真实多会话并行才成立。
## 判定方法
- 以本轮实际可用的工具与会话能力为准,探测一次即可,不在每个动作前重复怀疑。
- 下表只是安装参考;当前环境工具列表与表格冲突时,以工具列表为准。
- 不发明工具名;缺失即降级,不尝试用别的工具冒充原语。
## 降级规则
| 缺失能力 | 降级行为 |
| --- | --- |
| `C4` | `briefbound-thread-coordination`、`briefbound-multi-agent-orchestration`、`briefbound-autonomous-collaboration-loop` 不进入协议流程;冲突处理降级为当前 owner 单会话串行:先缩小 scope、调整顺序,共享文件在最后写入前重新拉取;不模拟 relay、outbox、接管或恢复债务。 |
| `C5` | 不使用 `agent_coordination.py`;ownership、隔离与合并判定用 Git 等价证据(分支归属、worktree 列表、clean 状态、merge-base 吸收证明)。 |
| `preflight --write-kind` | 以等价 Git 检查替代:根 `main/master` 上的 development 写入仍必须先建 task worktree;integration 仍要求目标 clean 且交付已验证。 |
| `C3` | 独立可并行工作由当前 agent 串行完成,顺序按依赖;不为并行而并行。 |
## 各 harness 安装速览
安装目标与激活块注入由 `scripts/install_codex_library.py` 的 `--agent` 参数决定:
| harness | skills 目录 | 激活块指令文件 |
| --- | --- | --- |
| Codex CLI/App | `~/.codex/skills` | `~/.codex/AGENTS.md` |
| Grok Build | `~/.grok/skills` | `~/.grok/AGENTS.md` |
| Claude Code | `~/.claude/skills` | `~/.claude/CLAUDE.md` |
| ZCode | `~/.zcode/skills`(项目级 `.agents/skills` 亦可) | `~/.zcode/AGENTS.md`(项目级 `AGENTS.md`) |
| Cursor | `~/.cursor/skills` | 项目级 `AGENTS.md`(手动放置) |
| Gemini CLI | `~/.gemini/skills` | `~/.gemini/GEMINI.md` |
| OpenCode | `~/.config/opencode/skills` | `~/.config/opencode/AGENTS.md` |
能力现状(以本机实测为准):Codex App 具备 `C4`/`C5` 全量;其余 harness 一般只具备 `C1`–`C3`(部分提供子代理)。在这些环境中协作族技能按上表降级,Briefbound 其余 owner 技能不受影响。
## 维护规则
- 技能正文引用 `C4`/`C5` 原语时,必须处于本文件降级门槛覆盖之下(在所属 SKILL.md 显式声明门槛,或仅在协作族协议文档内出现)。
- 新增 harness 支持时同步更新本表与安装器 `--agent` 说明;安装实测后再移除“未实测”标注。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |