n-skills Marketplace
One marketplace for every AI agent. Write skills once, run them anywhere.
Evidence shows a plugin marketplace, but no specific implementation or documentation for permission management, user confirmation, data flow transparency, sensitive data handling, or dependency security. External effects (like auto-sync) exist but their impact is not explained. Rollback mechanism not mentioned. Source attribution has partial evidence: README and sources.yaml mention preserving .source.json and attribution, but actual file content not provided. Thus most criteria are 0, only source_attribution is 1.
Self-consistency: README claims support for multiple agents, but actual skill files not provided, cannot verify consistency, only partial evidence. Dependency availability: depends on yaml and GitHub Actions, but no locked versions or integrity checks, only partial. Failure messages: no error handling or failure message documentation, 0.
Audience and scenarios: README clearly targets developers, lists multiple agents and installation methods, evidence sufficient, 2. Capability boundaries: describes skill categories and inclusion criteria, but not specific capability boundaries for each skill, 1. Trigger precision: no trigger mechanisms or precise instructions provided, 0. Environment fit: lists compatibility with multiple agents, but no actual tests or configuration details, 2.
Information architecture: README well-structured with table of contents and categories, 2. Install notes: provides multiple installation commands, 2. Naming stability: skill names consistent, but no version history, 1. Examples and FAQ: installation examples present, but no FAQ, 1. Known limitations: none mentioned, 0. License: Apache-2.0 complete, 2. Versioning and changelog: package.json has version, but no changelog, 1. Maintenance responsibility: README mentions maintenance and sync, but not clear responsible party, 1.
Output usability: skill descriptions and install commands, but no actual output examples, 1. Marginal value: as a marketplace, provides aggregation value, but not proven actual utility, 2. Cost-benefit: simple installation, but no performance or resource consumption info, 1.
Claim traceability: claims in README have some links, but no concrete evidence, 1. Cross-source corroboration: no other sources to verify, 0. Fact-inference separation: partially distinguished, but not explicit, 1.
- The repository does not provide actual skill file contents, so the authenticity and security of skills cannot be verified.
- The auto-sync mechanism may introduce external code, but no security review process is described.
- Publisher identity is unverified, and source attribution is only partially evidenced.
What does this agent do, and when should you use it?
n-skills is a curated marketplace for AI agent skills, working with Claude Code, Codex, and openskills. It champions a universal approach using the SKILL.md format and AGENTS.md discovery file, enabling 'write once, run everywhere.' The repo includes native skills (orchestration, open-source-maintainer, gastown, zai-cli) and synced external skills (dev-browser). Installation is via Claude Code's plugin system or the openskills universal installer.
n-skills offers 5 skills: orchestration (multi-agent orchestration with cc-mirror tasks and TodoWrite), open-source-maintainer (end-to-end GitHub repo maintenance), dev-browser (browser automation with persistent page state), gastown (multi-agent orchestrator for Claude Code + Opus), and zai-cli (Z.AI vision, search, reader, and GitHub exploration via MCP). Users add the marketplace to Claude Code via /plugin marketplace add, or install universally via openskills. External skills are auto-synced daily from their source repos via GitHub Actions, preserving attribution via .source.json.
- A developer using Claude Code wants to quickly install an orchestration skill to manage multi-agent tasks.
- An open-source maintainer needs an end-to-end repo maintenance skill to automate common chores.
- A developer automating browser tests wants persistent page state to simplify flows.
- A Codex user installs zai-cli to get vision and search capabilities.
- A Cursor or Windsurf user wants a unified skill installation experience via openskills.
What are this agent's strengths and limitations?
- Universal skill format (SKILL.md + AGENTS.md) works across multiple agents, reducing vendor lock-in.
- Curated marketplace ensures high-quality, real-value skills, avoiding low-effort wrappers.
- External skills auto-sync daily, keeping them in line with upstream repos.
- Diverse skill set covering workflow, tools, and automation categories.
- Limited number of skills (5), which may not cover all needs.
- Relies on AGENTS.md and openskills ecosystem; agents not supporting these may require extra configuration.
- Some skills (e.g., gastown) may require specific models (Opus) for optimal performance.
- As a marketplace, it needs ongoing maintenance and curation to stay relevant.
How do you install or deploy this agent?
Option 1: Claude Code plugin — run /plugin marketplace add numman-ali/n-skills, then /plugin install <skill>@n-skills for each skill. Option 2: Universal via openskills — npm i -g openskills, then openskills install numman-ali/n-skills and openskills sync. Requires Node.js.
How do you use this agent?
After adding the marketplace, install individual skills with commands like /plugin install orchestration@n-skills. For openskills, run openskills sync to activate. Each skill folder contains a SKILL.md that the agent reads to understand the skill.