Trust Engine/Software Library

taniwhaai/arai

Your AI assistant reads CLAUDE.md and ignores it anyway. Arai makes instruction files enforced — hooks that block, a tamper-evident audit trail, per-rule compliance stats.

6 starsRustApache-2.0ai-guardrailsclaude-codeclicode-quality
REVIEWED85/100

Scanned 2026-07-07 · deepseek v1

Dimension breakdown

Security

72/100

6 findings

Quality

88/100

No findings

License

95/100

No findings

Completeness

95/100

No findings

Findings (6)

MEDIUM: 3LOW: 2INFO: 1
  • MEDIUMsecurityInstall script uses curl | sh without mandatory checksum verification
  • MEDIUMsecurityAudit log hash chain uses SHA-256, not a memory-hard function
  • MEDIUMsecurityTelemetry endpoint allows plain HTTP for loopback
  • LOWsecurityNo explicit input validation on rule extraction from instruction files
  • LOWsecurityMCP authentication uses constant-time compare but no rate limiting

Detailed file:line citations + recommended fixes are visible to the maintainer of this repo after claiming the listing.

Maintainer of this repo?

Claim this listing in 30 seconds with GitHub OAuth. You'll see detailed findings, get notified about matched bounties, and can trigger re-scans on demand.

Claim listing

Embed the badge.

Drop it into your README. Every view backlinks to this verification page.

Trust badge — flat styleTrust badge — for-the-badge style

Markdown

[![Archimedes Trust](https://archimedes.market/api/badge/taniwhaai/arai.svg)](https://archimedes.market/r/taniwhaai/arai)

HTML

<a href="https://archimedes.market/r/taniwhaai/arai"><img src="https://archimedes.market/api/badge/taniwhaai/arai.svg" alt="Archimedes Trust Verified"/></a>