CommandVault is a library of reusable instructions for AI tools: prompts, project context files (CLAUDE.md, AGENTS.md, Cursor rules), Claude Code commands and subagents, Agent Skills and MCP servers. Each entry has fill-in fields, so you copy a finished version for your project instead of editing placeholders.
Who runs it
CommandVault is built and operated by Cacele Digital, a software studio in St. Augustine, Florida. The editor is William Geronco. Entries marked CommandVault editors were written or curated by our team. Members can also post their own.
Questions, corrections or takedown requests: contact us.
Where entries come from
- Original: written by our team or a member for CommandVault.
- Adapted:based on someone else's work, with credit and a link to the source.
- Directory listing:a skill or MCP server made by someone else. We write the description and install steps; the software itself installs from its author's repository, under its own license. We never copy it.
We use AI tools to help draft some descriptions and templates. An editor checks an entry before it's marked Reviewed, and nothing is marked Tested without a real run that we record with the tool, version and date.
How we review
- Reviewed:an editor checked that the entry does what it says, that every fill-in field is used, that install steps match the tool's current documentation, and that the labels are right.
- Tested: we ran it in the named tool and version on the shown date. The page includes what we tried and what happened.
- Not yet reviewed:published, but an editor hasn't checked it yet. Treat it as a starting point.
Votes, comments and uses come from real visitors only. We don't buy or create votes, accounts or reviews. Entries with problems can be reported from their page; we fix, merge or retire them.
What the labels mean
Prompts and templates
- Low risk: produces text or advice; nothing runs until you choose to.
- Medium risk: guides changes to code, data or settings. Review before you apply.
- High risk: touches security, money, production systems or personal data. Test somewhere safe first.
MCP servers and skills
These describe what the tool can do once installed, not whether its publisher is trustworthy.
- Read-only: Looks things up; it can't change your accounts or files.
- Can change your data: Has tools that create, edit or delete things in the connected service.
- Can move money: Can create payments, invoices or refunds. Use test mode first.
- Runs commands: Acts on your computer, browser, database or cloud with your credentials.
Whatever the label, read the tool list and give it the narrowest access that works, ideally a test account first.
Ranking
Lists are ordered by community votes, real use (copies, in-page runs and fetches through our MCP server) and freshness. While the community is small, we also feature workflows our editors have tested.