Skip to content

Latest commit

 

History

History
54 lines (38 loc) · 1.72 KB

File metadata and controls

54 lines (38 loc) · 1.72 KB

Roadmap

Vision

ai-rsk must become an impassable wall. The AI has no room to maneuver: it follows the rules or the build fails. The goal is that the AI says "I cannot do anything more than follow your instructions." ai-rsk scans itself: ai-rsk passes its own scan at 100/100.

Current: v0.7.1

  • 67 YAML rules (JS/TS, Python, Go, HTML)
  • Tree-sitter AST filter (Python, JavaScript/TypeScript)
  • 4 external tools (Semgrep, Gitleaks, osv-scanner, knip)
  • 5 compliance profiles (security, GDPR, AI Act, SEO, accessibility)
  • Security Score (0-100)
  • Default command (ai-rsk = ai-rsk scan)
  • Self-update (ai-rsk update)
  • Self-compliant (PASS 100/100 on itself)

v0.8.0: Go deep coverage

  • Issue #3: Gorilla, Gin, Echo, standard library patterns
  • Go-specific Layer 1 rules

v0.9.0: Rust rules + Tree-sitter improvements

  • Issue #4: Rust rules (unsafe, panic, unwrap, todo, allow dead_code)
  • Issue #5: Tree-sitter AST layer improvements (prerequisite for anti-evasion)

v1.0.0: Quality gate + Anti-evasion

  • Issue #6: SBOM generation CycloneDX/SPDX (Cyber Resilience Act compliance)
  • Issue #7: Gamedev profile
  • Issue #8: Desktop profile
  • Issue #9: Mobile profile
  • Issue #16: ANTI-EVASION (critical objective)
    • String obfuscation detection
    • ai-rsk-ignore abuse detection
    • Variable indirection / dynamic property access
    • Dataflow analysis for semantic equivalents
    • The AI cannot bypass ai-rsk

v2.0.0: Runtime + Governance

  • Issue #10: Runtime monitoring / drift detection
  • Issue #11: Agent control plane
  • Issue #12: Data governance / privacy flows

Documentation

  • Issue #14: Wiki FAQ
  • Issue #15: Manifesto "Why ai-rsk exists"

Decision

v1.0.0 = the AI can no longer bypass ai-rsk. Non-negotiable.