Database Sentinel plugin
A Claude Code plugin that audits Supabase and MongoDB security with a score, ready fixes, a read-only MCP server and a CLI agent.
- 50 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The audit reads database schema and policies through a connection string, and schema and query results go to the chosen model.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/database-sentinelDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The audit reads database schema and policies through a connection string, and schema and query results go to the chosen model.
This is third-party code. Review the repository files before installing.
What it does
Database Sentinel answers "audit my database" with a scored report and exact fix code. It works three ways: as a Claude skill (Supabase and MongoDB), as a read-only MCP server `sentinel-mcp` for any AI client with four tools and an `audit` prompt, and as a CLI agent `sentinel-audit` on your own model for Supabase. Checks follow a catalog of anti-patterns: queries in `audit-queries` and separate fix hints. For Supabase it checks RLS policies, grants, functions and roles, and for MongoDB it describes introspection-based self-checks. The agent runs on any OpenAI-compatible model (OpenRouter, OpenAI, local Ollama), confirms findings with an anonymous probe, and the `--fix` flag only prints fix SQL and never executes it. The repository holds a bench of 25 labeled cases to measure accuracy.
Who it is for. Developers building apps on Supabase or MongoDB with agents who want to check they did not leave open tables, wrong policies or excess rights.
Good fit when
- A todo app was shipped with RLS off and tables are readable by anonymous users
- An update policy without value checks lets users change others rows and this must be found
- A database check is embedded into CI with a free rules-only set
- Your own agent should get read-only database access for an audit
- You want ready SQL fixes that you will run yourself
Not a fit when
- You need automatic repair without review: fixes are only printed and you run them
- The database is not Supabase or MongoDB: other databases are not covered
- You would have to pass a connection string with broad rights: the audit needs a separate read-only role
Example request
Audit my Supabase database: find tables without RLS and overly broad grants, show the score and fix SQL without running it.Limitations
For Supabase create a separate read-only role with `BYPASSRLS`, and `sentinel-mcp --role-sql` prints the role SQL. You need `uv` to run the MCP server. The agent sends schema and query results to the chosen model, so use a local one for private projects. MongoDB has only the skill, with no MCP or agent. On the held-out split the agent F1 is about 0.83 and through MCP 0.87, so false alarms are possible and the report should be read critically. The author also offers a paid hosted version, but the repository works on its own.
How to disable. Disable the `database-sentinel` plugin through `/plugin` or delete the skill folder from `~/.claude/skills/`. Revoke the read-only role in Supabase in the database itself and remove the CLI with `pipx uninstall database-sentinel`.
FAQ
How do I install Database Sentinel in Claude Code?
Run `claude plugin marketplace add Farenhytee/database-sentinel`, then `claude plugin install database-sentinel@database-sentinel`. `uv` is required. Then set the connection with `/plugin configure database-sentinel@database-sentinel`.
Does the tool change my database?
No. The MCP server is read-only and the CLI `--fix` flag prints fix SQL and never runs it.
How do I connect it to Cursor?
Through the Add to Cursor button in the README or the command `uvx --from git+https://github.com/Farenhytee/database-sentinel sentinel-mcp` in the MCP configuration.
Which model does the CLI agent use?
Any OpenAI-compatible one: OpenRouter, OpenAI or local Ollama. The standard audit uses one prompt with an anonymous probe and `--deep` runs the tool-using agent.
What is covered for MongoDB?
The skill with anti-patterns, introspection, fix templates and a test recipe. There is no MCP server or CLI agent for MongoDB.
Discussion
No questions or reviews yet.
Related
PluginsA code security plugin: finds vulnerabilities, challenges its findings, proves exploitability and fixes with a verified patch
An MCP server for security document review: an agent submits files and the findings stay drafts until a person decides
A static scanner for vulnerabilities in AI agent code and MCP configs with 66 rules mapped to OWASP Agentic Top 10
A CLI and MCP that find RLS gaps, service_role keys in client code and unsigned Stripe webhooks in an app on Vercel and Supabase
A local MCP server and CLI find vulnerabilities in AI-generated code and drive a scan, fix and rescan loop, no account needed
Security Detections MCP: a detection rule database
Security Detections MCP
An MCP server that merges Sigma, Splunk, Elastic and KQL rules into one database for finding detections and ATT&CK gaps
README in short
The README keeps a changelog with benchmark results, then shows a quick start for Claude Code, Cursor and other MCP clients, skill and CLI agent installation and MCP usage.