# Skill Store Marketplace: A Security-Audited Registry for Claude and Codex Skills > Skill Store is an open-source repository of vetted agent skills for Claude, Codex, and Claude Code, with automated security scans and a one-command CLI installer. The audit is report-only, which is both a strength and a limitation. **aiskillstore/marketplace** — Security-audited skills for Claude, Codex & Claude Code. One-click install, quality verified. - Repository: https://github.com/aiskillstore/marketplace - Website: https://skillstore.io - Stars: 443 · Forks: 48 - Language: Python - License: not declared - Published: 2026-08-27 · Updated: 2026-08-27 · Language: en - Canonical page: https://hysenlabs.com/projects/aiskillstore-marketplace ## A Registry That Automates the Skill Supply Chain Skill Store solves a specific problem: finding agent skills that are not just functional but also vetted for dangerous patterns. The repository is the content layer behind skillstore.io, storing every approved Agent Skill, its metadata, and the automated security audit published alongside it. The target user is an engineer who wants to install a skill into Claude Code or Codex without manually reviewing every line of code. The README is clear that this is not a place to submit skills via pull request; the platform's review pipeline writes to this repo automatically. That separation matters: the repository is a mirror of a curated process, not a community dump. ## How the Review Pipeline and Audit Interact The submission flow starts at skillstore.io/submit, where a GitHub repository URL containing a SKILL.md is entered. The system then runs automated security analysis before a maintainer reviews and approves it. On approval, the skill is published to this repo and appears on the site. The audit scans for dangerous code patterns like eval and exec, raw system commands, file access outside the project scope, network calls to external hosts, obfuscated or minified code, and credential handling. Here is the key design choice: the security analysis is report-only. Findings inform maintainers and users, but a risk result does not automatically block publication. That means the badge you see is a disclosure, not a certification. The README points to a Security Trust Center for methodology and limitations, which is the right place to look before trusting any skill. ## One-Command Install for Claude Code and Codex Installation is the most concrete part of the project. The recommended path is the skillstore CLI, run via npx: npx skillstore add author/skill-name. The example given is npx skillstore add aiskillstore/code-review. The CLI downloads the skill and places it in the correct skills directory for your tool. For Claude Code, auto-discovery happens; for Codex, you need to restart the session. The README also mentions manual installation and ZIP upload for Claude Web, with install guides covering scope directories like ~/.agents/skills/, .claude/skills/, ~/.claude/skills/, and .codex/skills/. This is a real advantage over fetching skills from a raw GitHub repo yourself, because the CLI handles the path mapping that varies by tool. ## Repository Layout: Where the Automation Lives The repo has a clear structure: skills/ holds approved published skills, pending/ holds submissions awaiting review, packages/cli contains the skillstore CLI, packages/skillstore holds the core package, schemas/ defines JSON schemas for skill records, and scripts/ contains maintenance and scoring scripts. The .github/workflows directory runs submission, audit, and sync automation. This layout shows that the project is not just a static list; it is a pipeline with machine-readable schemas and CI workflows. The schemas are worth noting because they let you validate skill records programmatically, which is useful if you build tooling on top of this catalog. ## The Limitation: Report-Only Audits and Manual Review The most significant limitation is that a skill can be published even if the automated audit flags a risk. The README states that a risk result does not automatically block an approved skill. That means the 'security-audited' tag in the description is softer than it sounds. A skill with an eval call could still appear, as long as a maintainer approves it. The other limitation is that the repo is not a submission channel. If you want to contribute a skill, you must go through the web platform, and pull requests adding skills are closed. For an engineer who wants to fork the repo and add a skill locally, that is a dead end. The audit also cannot catch everything; the README does not claim it can, but a user scanning the description might assume a higher level of safety than the report-only model delivers. ## Alternatives: Manual Installation and Raw GitHub The obvious alternative is to install skills directly from a GitHub repository by copying the SKILL.md and its supporting files into the appropriate skills directory yourself. That approach gives you full control over what you install, but it also means you are responsible for reviewing the code. Another alternative is to use the Agent Skills specification directly from agentskills.io and build your own skills from scratch. The difference in approach is trust: Skill Store centralizes the review process and gives you a security passport, while manual installation puts the burden on you. For a single skill, manual install is fine; for a team that needs a repeatable process, the CLI and the registry are more convenient. The trade-off is that you are trusting Skill Store's audit methodology, which you should read before relying on it. ## Maintenance, Licensing, and Upgrade Cost The repository is actively maintained, with releases like skillstore-v0.1.13 pushed on 2026-07-22, and the last push on the same date. The CLI is versioned, so upgrading is a matter of running npx again, which pulls the latest version. The marketplace catalog is MIT-licensed, so the repository metadata is freely reusable. However, individual skills carry their own licenses, and the README explicitly says to check each skill's LICENSE file. That is a real upgrade cost: before adopting a skill, you must verify its license separately, because the MIT license on the catalog does not extend to the skills themselves. There is no mention of a deprecation policy for skills, so a skill could be removed or abandoned without notice. The automation that maintains the repo means manual changes are limited to maintainers, so you cannot fix a skill's bug by editing this repo. ## Conclusion Adopt Skill Store if you want a central, quality-checked source of agent skills for Claude Code or Codex, especially if you value automated security scanning before installation. Avoid it if you need a hard security guarantee, because the audit is report-only and risk findings do not block publication. Before adopting, verify the specific skill's LICENSE and review the Security Trust Center methodology, and test that the skill's network and file access patterns match your project's trust boundary. ## FAQ ### What is the Skill Store marketplace repository? It is the open-source content repository behind skillstore.io. It holds every approved Agent Skill, the records that go with each one, and the automated security audits published alongside them, all written by the platform's review pipeline. ### How do I install a skill from Skill Store? Run `npx skillstore add author/skill-name`, for example `npx skillstore add aiskillstore/code-review`. It downloads the skill into the right `skills/` directory for your tool. Claude Code auto-discovers it, while Codex needs the session restarted. ### How do I submit a skill to Skill Store? Through skillstore.io/submit, not through a pull request. You give the GitHub repository URL containing the `SKILL.md`, the submission goes through automated security analysis, a maintainer reviews and approves, and only then is it published into this repository. Pull requests adding skills are closed. ### Does a failed security audit stop a Skill Store skill from being published? No. The scan flags dangerous code patterns, file access outside project scope, network calls to external hosts, obfuscated code and credential handling, but the analysis is report-only. Findings inform maintainers and users, and a risk result does not automatically block an otherwise approved skill. ### What does Skill Store require for a valid skill? A `SKILL.md` is required per the Agent Skills specification, supporting files the skill references are optional, and a `LICENSE` is recommended. ### What licence applies to skills from Skill Store? The catalogue itself is MIT-licensed, but individual skills carry their own licences and you are expected to check each skill's `LICENSE` file. The three most recent CLI releases, v0.1.11, v0.1.12 and v0.1.13, all shipped on 2026-07-22. ## Sources - [Official documentation](https://skillstore.io) - [Official README](https://github.com/aiskillstore/marketplace#readme) - [Project repository](https://github.com/aiskillstore/marketplace) - [Release notes](https://github.com/aiskillstore/marketplace/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/aiskillstore-marketplace