Commonly Misused or Unsafe Patterns

This page catalogs anti-patterns that conflict with AegisLayer governance principles.

Anti-Patterns

• Implicit authority inferred only from prompt text
• Connector invocation without explicit capability scope
• Policy evaluation after external execution
• Reusable approvals without scope or expiry binding
• Silent fallback from denied state to broader privilege path
• Evidence records without stable identifiers or timestamps
• Trust monitoring without enforcement actions

Safer Replacements

• Identity-first validation with explicit authority context
• Fail-closed connector readiness checks
• Policy-before-execution control point
• Request-bound and time-bound approvals
• Immutable finalized evidence bundles

Related:

• Pattern Library
• Security Design Principles