llm-safety-patterns

SKILL.md

Identifiers flow AROUND the LLM, not THROUGH it. The LLM sees only content. Attribution happens deterministically.

| userid | UUID | Can be hallucinated, enables cross-user access | | tenantid | UUID | Critical for multi-tenant isolation | | analysisid | UUID | Job tracking, not for LLM | | documentid | UUID | Source tracking, not for LLM | | artifactid | UUID | Output tracking, not for LLM | | chunkid | UUID | RAG reference, not for LLM |

| sessionid | str | Auth context, not for LLM | | traceid | str | Observability, not for LLM | | Any UUID | UUID | Pattern: [0-9a-f]{8}-... |

Security patterns for LLM integrations including prompt injection defense and hallucination prevention. Use when implementing context separation, validating LLM outputs, or protecting against prompt injection attacks. Source: yonatangross/orchestkit.

View raw