zjs105910/workbuddy-market / bundled/github-project-manager/skills/github-project-manager
github-project-manager
Verified管理从 GitHub 下载到本机的开源项目与技能包:列出、扫描发现、安装前预检、检查更新、一键更新、移除、回滚、数据完整性校验与环境诊断。带 O_EXCL 原子锁 + 心跳租约、事务级失败回滚、InstallPlan 状态收敛、单次遍历的磁盘预检、批量更新复用远程信息(消灭 N+1 请求)、ZIP 预扫描与压缩炸弹/设备名双重防护、skill 冲突阻断、内容指纹校验、硬链接与 rename 两种备份介质。触发方式:/ghpm、『GitHub 项目管理』『检查插件更新』『更新开源项目』『我装了哪些 GitHub 项目』『这个插件有新版吗』『卸载插件』『这个仓库能装吗』『ghpm doctor』『ghpm verify』『文件丢了/装坏了』。
scripts/ghpm/discovery.py
View on GitHub# -*- coding: utf-8 -*-
"""skill 发现:从 ZIP 索引或本地目录里认出 ``<任意深度>/skills/<name>/SKILL.md``。
v4 的 index_source_tree() 把「找 skill」和「量体积」并进**同一次**遍历:
v3.2 find_skill_dirs() rglob("*.md") ← 遍历 1
path_size_bytes() rglob("*") ← 遍历 2
每个 skill 子目录再各 rglob("*") ← 遍历 3..N
v4 index_source_tree() 一次 scandir 递归,同时产出技能清单 / 全树体积 / 各技能体积
顺带把 os.scandir 换成底层调用:DirEntry 自带 stat 结果,
比 ``rglob + f.stat()`` 少一次系统调用 / 每项。
"""
from __future__ import annotations
import re
from pathlib import Path
from .models import SourceIndex, TreeStats
from .util import iter_tree
def match_skill_path(parts) -> str | None:
"""判断是否 <任意深度>/skills/<name>/SKILL.md,是则返回 skill 名。
注意仓库里的 skills 目录不一定在顶层:oh-story-dsh 就是
packages/knowledge/oh-story/skills/<name>/SKILL.md。"""
if len(parts) < 3:
return None
if parts[-1].lower() != "skill.md":
return None
if parts[-3].lower() != "skills":
return None
name = parts[-2]
return name or None
def skill_name_from_meta(d: Path, fallback: str) -> str:
"""单 skill 仓库用 SKILL.md frontmatter 里的 name,没有就用仓库名"""
f = Path(d) / "SKILL.md"
try:
txt = f.read_text(encoding="utf-8", errors="replace")[:2000]
except Exception:
return fallback
m = re.search(r"^name:\s*[\"']?([A-Za-z0-9_.-]+)[\"']?\s*$", txt, re.M)
return (m.group(1) if m else fallback) or fallback
def index_source_tree(root, fallback: str = "") -> SourceIndex:
"""一次遍历拿到:技能清单 + 整树体积 + 每个 skill 目录的体积。
技能识别规则与 v3.2 find_skill_dirs 完全一致(包括「按路径排序先到先得」
和「根目录单个 SKILL.md 算单 skill 仓库」两条),只是合并到同一次遍历里。
"""
root = Path(root)
stats = TreeStats()
skill_mds: list = []
# 目录 → 该目录下**直接**文件的字节数;比记每个文件省内存(大仓库 20 万文件时差一个量级)
dir_file_bytes: dict = {}
for d, e in iter_tree(root):
try:
if e.is_dir(follow_symlinks=False):
stats.dirs += 1
continue
if not e.is_file(follow_symlinks=True):
continue
size = e.stat(follow_symlinks=True).st_size
except OSError:
continue
stats.files += 1
stats.bytes += size
dir_file_bytes[d] = dir_file_bytes.get(d, 0) + size
if e.name.lower() == "skill.md":
skill_mds.append(Path(e.path))
out: dict = {}
for f in sorted(skill_mds): # 与 sorted(root.rglob("*.md")) 同序
try:
parts = f.relative_to(root).parts
except ValueError:
continue
name = match_skill_path(parts)
if name and name not in out:
out[name] = f.parent
if not out and (root / "SKILL.md").is_file():
out[skill_name_from_meta(root, fallback or root.name)] = root
skills = [(k, out[k]) for k in sorted(out)]
skill_bytes = _bucket_by_skill(root, out, dir_file_bytes)
return SourceIndex(root=root, stats=stats, skills=skills, skill_bytes=skill_bytes)
def _bucket_by_skill(root: Path, out: dict, dir_file_bytes: dict) -> dict:
"""把「每个目录自己的文件字节」归到它所属的 skill 名下。
沿祖先链上溯(深度有限),命中 skill 目录就归它 —— O(目录数 × 深度),
比「对每个 skill 再遍历一遍子树」少一个量级。
"""
skill_dir_to_name = {p: k for k, p in out.items()}
result = {k: 0 for k in out}
for d, size in dir_file_bytes.items():
cur = d
for _ in range(64):
name = skill_dir_to_name.get(cur)
if name is not None:
result[name] += size
break
if cur == root:
break
parent = cur.parent
if parent == cur:
break
cur = parent
return result
def find_skill_dirs(root: Path, fallback: str = ""):
"""在本地解压目录里找 skill,返回 [(name, dirPath)]。
两种形态都认:① 多 skill 包 .../skills/<name>/SKILL.md
② 单 skill 仓库 —— 根目录直接一个 SKILL.md
保留这个入口是为了兼容既有调用方;实现已并进 index_source_tree。
"""
return index_source_tree(root, fallback).skills
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: 470k | 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 |