E01¶


Research	R0021 — Prompt engineering definitions
Run	2026-03-25
Query	Q006
Source	SRC01
Evidence	SRC01-E01
Type	Reported

RFC 2119 applied to AI coding agent specifications

URL: https://deliberate.codes/blog/2026/writing-specs-for-ai-coding-agents/

Extract¶

Marco Naetlitz (Feb 2026) applies the full RFC 2119 vocabulary to spec-driven development with AI coding agents:

SHALL/MUST — absolute requirements (e.g., "it SHALL return the computed numeric result")
SHOULD — strong recommendations (e.g., "it SHOULD indicate the position of the error")
MAY — optional behaviors (e.g., "it MAY suggest corrections for common mistakes")
SHALL NOT — explicit prohibitions (e.g., "it SHALL NOT retry automatically to avoid account lockout")

The author argues: "Using only SHALL forces you to either make everything mandatory or omit optional behaviors entirely. The full RFC 2119 vocabulary lets you express degrees of importance."

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Supports	Published example of RFC 2119 applied to AI specifications
H2	Contradicts	At least one published example exists
H3	Supports	This is AI agent spec-writing, not mainstream prompt engineering

Context¶

This is the only published example found of RFC 2119 applied specifically to AI behavioral specifications. It is a practitioner blog post from 2026, not a formal standard.