菜单
Router contract for provider-aware Gobby help and installed skill dispatch.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Instructions for using gcode CLI for code search and retrieval. Loaded on demand when project has a code index.
Pre-transition checklist for closing or reviewing tasks. Covers fresh verification evidence, commits, and memory gates.
Use when processing CodeRabbit review comments, CodeRabbit CLI reports, or `$gobby coderabbit [findings]` requests.
Use when creating, editing, or validating Gobby skills, especially when a skill must change agent behavior or become discoverable through gobby-skills.
Use when coordinating a full gobby build run for an epic or task, especially when the user assigns the current session as coordinator, asks for a coordination epic, wants build agents/worktrees monitored, or wants gobby build bugs fixed so future runs work unattended.
Select the best bundled developer agent for automated expansion leaves and document fallback decisions.
| name | gobby |
| description | Router contract for provider-aware Gobby help and installed skill dispatch. |
| version | 2.0.0 |
| category | core |
| triggers | help |
Gobby skill routing is provider-dependent. Codex uses $gobby; providers with
an installed slash router use /gobby. The router advertises installed skills
on bare help requests and routes named skill requests through gobby-skills.
For Codex help requests ($gobby, $gobby help) and slash-router help
requests (/gobby, /gobby help), show dynamic help generated from installed
skills. Do not maintain a hand-written shortcut list.
Use list_skills on gobby-skills when skill discovery is needed. Present
user-invoked skill examples with the provider's active trigger:
$gobby <skill>/gobby <skill>Do not present /gobby as universal syntax.
These forms route to get_skill(name="<skill>") on gobby-skills:
$gobby <skill> [args]
$gobby skill <skill> [args]
/gobby <skill> [args]
/gobby skill <skill> [args]
/gobby:<skill> [args]
The router emits a fetch directive only. It does not inline skill bodies. Preserve the user's trailing arguments and continue after the named skill is loaded.
For MCP tool access, use progressive discovery:
list_mcp_servers() — discover serverslist_tools(server_name="...") — discover toolsget_tool_schema(server_name="...", tool_name="...") — get parameterscall_tool(server_name="...", tool_name="...", arguments={...}) — execute