AI-SPM vs Karpathy Skills

AI-SPM (AI Security Posture Management) is an open-source control plane for AI agent security in production environments. Built by indie developer dshapi and posted to Hacker News, it addresses a real gap: most LLM systems now have tool access and decision-making power, but almost no runtime oversight layer to catch when things go wrong. The system works as a gateway between your application and the LLM, enforcing three main controls: prompt injection detection (including obfuscated variants that bypass naive pattern matching), structured tool call validation against defined policies using Open Policy Agent (OPA), and sensitive data leakage prevention (PII and model output filtering). An Apache Kafka and Apache Flink streaming pipeline provides real-time audit trails and anomaly detection. The creator's key insight is that tool misuse — not model jailbreaks — is the primary risk vector in production AI agents. A rogue or compromised agent that escalates tool permissions or exfiltrates data through sanctioned channels is far harder to catch than a classic prompt injection. AI-SPM is early, minimal traction, and needs real-world stress testing. But as AI agent deployments mature from demos to production, runtime security tooling like this becomes non-optional.

Karpathy Skills is a CLAUDE.md plugin distilled from Andrej Karpathy's public observations on LLM coding pitfalls. Drop the single file into your project root (or install it as a Claude Code skill) and every Claude Code session starts pre-loaded with the four principles Karpathy identified as most commonly violated: think before writing, prefer simplicity, make only targeted changes, and close loops with explicit verification. The project has accumulated 1,450+ GitHub stars in under two weeks. The implementation is intentionally minimal — it's a structured system prompt, not a framework. Each principle is spelled out with concrete anti-patterns to avoid: no premature generation, no over-engineering simple tasks, no cascading refactors when a surgical fix suffices, no ending a session without verifying the goal was actually met. It's Karpathy's "Software 2.0" thinking applied to the agent workflow meta-layer. What makes this compelling isn't the technology — it's the curation. Karpathy has spent more time thinking about LLM behavior patterns than almost anyone outside the major labs. Packaging that into something installable in 30 seconds lowers the floor for teams who want more reliable agent outputs without extensive prompt engineering work.