惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

M
MIT News - Artificial intelligence
雷峰网
雷峰网
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Last Week in AI
Last Week in AI
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
阮一峰的网络日志
阮一峰的网络日志
月光博客
月光博客
博客园 - Franky
腾讯CDC
T
Tailwind CSS Blog
Recent Announcements
Recent Announcements
V
V2EX
N
Netflix TechBlog - Medium
量子位
Jina AI
Jina AI
Y
Y Combinator Blog
The GitHub Blog
The GitHub Blog
G
Google Developers Blog
爱范儿
爱范儿
博客园 - 叶小钗
D
Docker
MongoDB | Blog
MongoDB | Blog
D
DataBreaches.Net
T
The Blog of Author Tim Ferriss

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - AI-Colleagues/skill-mgr: Agent Skill manager. In...
NeuralNotwor · 2026-04-23 · via Hacker News: Show HN

CI Coverage PyPI

skill-mgr is a standalone multi-agent skill installer and inspector for AgentSkills-compatible SKILL.md packages. It installs one validated skill into bundled agent targets, supports local directories and GitHub shorthand refs, and exposes the same lifecycle surface as the Orcheo skill workflows: install, update, uninstall, validate, list, and show.

Installation

uv tool install skill-mgr

Run the installed CLI directly:

skill-mgr -h

If you want to use the CLI without installing it first:

uv run skill-mgr -h

If you are working from source:

uv sync --all-groups

Commands

skill-mgr install REF [--target TARGET ...] [--format rich|markdown|json]
skill-mgr update REF [--target TARGET ...] [--format rich|markdown|json]
skill-mgr uninstall NAME [--target TARGET ...] [--format rich|markdown|json]
skill-mgr validate REF [--format rich|markdown|json]
skill-mgr list [--target TARGET ...] [--format rich|markdown|json]
skill-mgr show NAME [--target TARGET ...] [--format rich|markdown|json]
skill-mgr support-matrix [--format rich|markdown|json]

Rules:

  • Repeated --target/-t values are allowed.
  • Omitting --target installs to bundled agents detected in the current environment.
  • Explicit --target values bypass detection and are still honored even if the agent home directory does not exist yet.
  • all is mutually exclusive with explicit targets.
  • Rich-rendered human output is the default.
  • Use --format markdown for plain-text Markdown tables that are easier for LLMs and other text consumers to parse.
  • Use --format json for strict structured output.

Source Refs

Supported REF forms:

  • Local skill directory path containing SKILL.md
  • GitHub shorthand owner/repo
  • GitHub shorthand owner/repo/path/to/skill

Resolution behavior:

  • Existing local paths always win after home expansion and normalization.
  • Only non-existent local refs fall through to GitHub shorthand parsing.
  • GitHub installs resolve the repository default branch via the GitHub API, download one tarball, safely extract it once, then reuse the materialized skill directory across all selected targets.
  • Nested GitHub paths must resolve to a directory inside the archive.

Initial OS Support Matrix

The bundled adapters currently publish this matrix:

Adapter Windows Linux macOS Managed install root Notes
claude supported supported supported ~/.claude/skills Home-relative managed skill root used by the adapter
codex supported supported supported ~/.codex/skills Matches the local Codex skill layout used by the app/CLI
openclaw supported supported supported ~/.openclaw/skills Matches OpenClaw's documented managed/local skills directory
orcheo supported supported supported ~/.orcheo/skills Matches Orcheo's managed local skills directory

Platform semantics:

  • supported: the adapter has a defined install root and is expected to work on that OS.
  • unsupported: the adapter is intentionally skipped on that OS.
  • unknown: the adapter root is not yet defined and the target is skipped with unknown_install_root.
  • agent_not_detected: the adapter is bundled, but its home directory was not found during default target selection.

Current bundled adapters all map to explicit home-relative roots. Explicit --target all expands to claude, codex, openclaw, and orcheo on Windows, Linux, and macOS, while the default target set only includes agents detected on the current machine.

SKILL.md Contract

validate enforces the following SKILL.md contract:

  • SKILL.md must exist at the resolved skill directory root.
  • The file must start with valid YAML frontmatter delimited by --- lines.
  • name is required.
  • name must be 1-64 characters of lowercase letters, numbers, and internal hyphens.
  • description is required and must be a non-empty string.
  • license, when present, must be a string.
  • compatibility, when present, must be a string.
  • metadata, when present, must be a mapping.
  • allowed-tools, when present, must be a string.

Recognized fields are normalized into the validation output:

  • name
  • description
  • license
  • compatibility
  • metadata
  • allowed-tools

Unknown frontmatter keys are preserved in extra_fields in the library result so downstream tools can retain adapter-specific metadata.

Minimal valid example:

---
name: demo-skill
description: Explain what the skill does and when to use it.
---

Skill instructions go here.

Examples

Install from a local directory into every bundled target:

skill-mgr install ~/skills/demo-skill

Install only for Codex and Claude:

skill-mgr install ~/skills/demo-skill -t codex -t claude

Install from a GitHub repo root:

skill-mgr install owner/repo

Install from a nested GitHub skill directory:

skill-mgr install owner/repo/skills/demo-skill

Validate without installing:

skill-mgr validate owner/repo/skills/demo-skill --format markdown

List skills across bundled adapters:

skill-mgr list

Development

make format
make lint
make test

The CI workflow currently runs formatting, linting, type-checking, and pytest on Windows, Linux, and macOS.