$ AI agent governance, security tooling, and mechanical enforcement.

rigscore v0.1.0

rigscore, security, ai, devtools, mcp, open-source

Why this exists

AI coding tools ship fast and configure loose. Claude Code, Cursor, Windsurf — they all read your filesystem, execute commands, and call APIs. Most developers set them up once and never look at the configuration again.

rigscore checks the things that matter and gives you a number. No accounts, no telemetry, no network calls. Run it, see the score, fix what’s broken.


What v0.1.0 checks

8 checks, each contributing to a score out of 100:

CheckWhat it scans
MCP server configurationTransport type, wildcard env passthrough, filesystem scope
CLAUDE.md governanceGovernance file existence, forbidden actions, approval gates
Secret exposure.env in .gitignore, API key patterns in configs, file permissions
Docker securitySocket mounts, privileged mode, host paths
Git hooksPre-commit hooks, push URL guards
Permissions hygieneSSH directory/key permissions, world-readable sensitive files
Skill file safetyInjection patterns, shell execution, external URLs
Container securityMissing user/cap_drop, host network mode

How it works

1
npx rigscore

Scans the current directory, evaluates each check, prints a score. JSON output available for CI integration with --json.

Every check runs locally. Nothing leaves your machine.


Install

1
npx rigscore

MIT licensed. github.com/Back-Road-Creative/rigscore

Configuration details reflect a production environment at time of writing. Implementation specifics vary based on tooling versions, platform updates, and organizational requirements. Validate approaches against current documentation before deployment.