danyuchn/asd-ste100-skill

What it solves

It addresses the problem of ambiguity in English text that AI agents must parse, such as tool descriptions, error messages, and inter-agent instructions. By applying the principles of ASD-STE100 Simplified Technical English—a standard originally designed for aerospace maintenance to prevent fatal misinterpretations—it ensures that downstream agents receive clear, literal, and unambiguous instructions without needing human clarification.

How it works

The project provides a skill for Claude Code that analyzes and rewrites text based on two modes: Strict (for procedures and errors) and STE-flavored (for prose like READMEs). It identifies rule violations—such as passive voice, complex tenses, ambiguous word choices, and overly long sentences—and rewrites them into a simplified form. It includes a deterministic linter to check for structural patterns like semicolons and specific phrasal verbs, while applying the general principles of the STE dictionary (using the plainest word consistently) without redistributing the restricted official dictionary itself.

Who it’s for

Developers building AI agent workflows where agents communicate with other agents, or where agents must interpret technical tool descriptions and system prompts without human intervention.

Highlights

• Dual-mode simplification: Offers a strict mode for critical instructions and a flavored mode for general documentation.
• Structural Linter: Includes a tool to mechanically check for forbidden patterns like nominalizations and marketing adjectives.
• Preservation of Precision: Rewrites text without dropping facts or scope qualifiers; if simplification would lose precision, it flags the trade-off instead.
• Agent-Centric Design: Specifically repurposes aerospace safety standards to improve machine-to-machine communication reliability.