simba-skills releases six guardrail skills for Claude Code

simba-skills has published six Claude Code skills designed to keep an AI coding agent from touching production. The skills block an agent from writing to live databases or sites during tests, repro loops, migrations and imports, and add guarded deploys, nightly backups and settings audits. Risky actions such as deploy, delete and settings changes only run when typed explicitly as /simba-skills:<name>. The repo ships an eval suite in which, according to simba-skills, 3 of 3 runs refused to test against a production .env with the plugin installed, versus 3 of 3 without it. Install is via the Claude Code plugin marketplace or npx skills@latest add simbaTmotsi/simba-skills; users should pick one method, not both, or every skill is installed twice. The plugin targets people running their own apps on a Linux server with Docker and Postgres while using an agent like Claude Code on the same machine; the server and deploy skills assume Docker Compose and Postgres and adapt to a project's existing deploy scripts.

Generated from reports · updated 56 min ago

Not enough continuous observations yet to draw a trend.