afrp

What problem does it solve? AI agents often generate misaligned responses because they skip intent classification and jump straight to output, producing answers calibrated to the wrong audience, wrong depth, or wrong interpretation. AFRP prevents this by enforcing a mandatory pre-routing gate that classifies every query before any execution begins. ## Core Features & Use Cases - Query Classification Gate: Classifies incoming queries on two axes — Type (A factual, B analytical, C implementation, D ambiguous) and Audience Level (Novice, Practitioner, Expert) — then maps them to the appropriate oramasys execution mode. - Intent-Verification Gate: Forces a stop-and-clarify check on interpretation risk, before negative conclusions, when explicit instructions conflict with assumptions, and when decisions depend on mutable external state like live PR or branch status. - Failure Mode Taxonomy: Ships an extended reference documenting ten failure modes (personalized slop, abstraction mismatch, handwaving, stale authority snapshots, and more) with recovery procedures. - Use Case: Before an agent answers "write guidance for small business owners on tariffs," AFRP classifies it as Type B for a novice audience, declares the scope explicitly, and calibrates the response to operational steps instead of abstract frameworks. ## Quick Start Ask the agent to run the AFRP gate on your query before generating any non-trivial response, for example: "Classify this request with AFRP, then answer it."