OWASP LLM-01: Direct & Indirect Prompt Injection Boundary

Statutory Framework: OWASP Top 10 for LLM Applications (LLM01:2025)
Rule ID: RULE-OWASP-001 | Severity: PROGRAMMATIC BLOCKER
Mitigation Action: SANCTIFY_INPUT_STREAM

1. Statutory Mandate & Legal Boundary

"Intercepts direct user prompt overrides and indirect RAG retrieval payload poisoning before model context window assembly."

2. Technical Invariant & AST Engine Matcher

[POLICIES.RULE-OWASP-001]
rule_id = "RULE-OWASP-001"
statute = "OWASP Top 10 for LLM Applications (LLM01:2025)"
severity = "BLOCKER"
action = "SANCTIFY_INPUT_STREAM"

[ENFORCEMENT]
fail_closed = true
audit_trail = "Ed25519_DAC"
latency_max_us = 800

3. Production Python / Rust Integration

from anchor import AnchorGuard

@AnchorGuard.guard(policy="RULE-OWASP-001")
def process_machine_payload(payload: dict):
    # Evaluated in <0.8ms against RULE-OWASP-001 AST query bounds
    return compliance_engine.enforce(payload)