360 Secure SkillHub
360 安全技能中心

doc-bridge-handoff

面向编码代理的仓库路由技能:在修改含 doc-bridge.config.json 的仓库前,先用只读 MCP 工具或 Node 解析脚本确定目标包,读取 readBeforeEditing 文档,将改动限定在 editRoots 内,并在宣称完成前执行 checks。解析器只给出指引,不授权编辑。

emersonbraun
emersonbraun
热度 7.65
最近更新 2026-08-12 07:37:36
分类:开发依赖:代码执行 / 读写文件 / MCP 服务应用场景:代码生成 / 工作流 / 任务管理

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add doc-bridge-handoff
Metadata
name
doc-bridge-handoff
description
Resolve Doc Bridge boundaries before changing code.
version
1.0.0
author
AgentsKit
license
MIT
platforms
["darwin","linux","windows"]
metadata
{"hermes":{"tags":["documentation","coding-agents","repository-routing"]}}

Doc Bridge handoff

Use this skill before editing a repository that contains doc-bridge.config.json.

  1. Resolve the package or ownership id that best matches the requested change:
    • Prefer the read-only MCP tool handoff.resolve when it is available.
    • Otherwise run node <skill-directory>/scripts/resolve-handoff.mjs <id> from the repository.
  2. Read every file in readBeforeEditing, beginning with startHere.
  3. Keep changes inside editRoots. If the requested path is not covered, stop and report the missing route instead of guessing.
  4. Make the smallest change that satisfies the request.
  5. Run every command in checks before claiming completion.
  6. If documentation changed, refresh the Doc Bridge index and run its gate.

If resolution fails, returns incomplete fields, or names an unknown target, stop. Do not infer edit permission from repository layout.

The resolver and MCP tools are read-only. They resolve project guidance but never authorize edits, publish changes, execute returned checks, or replace repository instructions. The skill uses no credentials and has no provider, hosted service, or AKOS dependency.

Portable runtimes

This directory follows the open Agent Skills layout: one SKILL.md plus optional scripts and fixtures. It can be loaded as a local skill by OpenClaw-compatible clients, Hermes Agent, Pi, Cursor, or another runtime that supports Agent Skills and shell execution. Runtime-specific publication metadata is intentionally kept outside the skill.