cyberneticlibrary

Detect and resist deceptive requests

anti-deceptionskillsetup L18
ejentum/ejentum-mcp
What it does

Detect inconsistencies and verify claims against evidence

Best for

Catches narrative contradictions invisible to humans reading sequentially—flags shifts in story that seem plausible alone.

Inputs
  • · narrative or claim
  • · available evidence/documents
Outputs
  • · contradiction list
  • · verification status per claim
  • · confidence scores
Requires
  • · vector search for evidence matching
  • · LLM for semantic consistency checking
Preconditions

Claims articulated clearly; evidence corpus indexed

Failure modes

False positives if evidence corpus incomplete; missed subtle contradictions if claims rephrased; high false negatives if evidence sparse

Trust signals
  • · side-by-side quote comparison for contradictions
  • · confidence scores based on evidence strength
  • · explicit gaps where claims lack supporting evidence