Aerospace has spent decades making maintenance manuals unambiguous with a controlled language called Simplified Technical English. STE-Prompt packages that discipline as a plugin for Claude Code, Codex, and any coding agent.
ASD-STE100 combines ~60 writing rules — short sentences, active voice, one instruction each — with a controlled dictionary where every approved word carries exactly one meaning. Ask a model to answer in STE and it mechanically drops the two things that make LLM prose hard to trust: hedging, and compound sentences that smuggle three claims into one.
"In order to deploy the service, you'll first want to make sure that all of the environment variables have been properly configured, and then you can go ahead and trigger a new build, which should subsequently pick up the latest changes."
Deploy the service:
1. Make sure you set all the environment variables.
2. Start a new build.
3. The build uses the latest changes.
This repo is a Claude Code plugin and a one-plugin marketplace. Add it straight from GitHub:
/plugin marketplace add menonpg/ste-prompt
/plugin install ste-prompt@menonlabThen rewrite anything on demand:
/ste rewrite the deploy steps above in Simplified Technical English
# or just ask — the skill triggers automatically/ste rewrites your last answer or any block into Simplified Technical English on demand.
Set it persistent so the agent writes every explanation and runbook in STE by default.
Append AGENTS.ste.md to your repo's AGENTS.md (or ~/.codex/AGENTS.md) — works in Codex, Cursor, Aider & more. The behavior travels with the code.
Built for procedures, runbooks, and agent output — not for nuance or risk analysis, where the caveats are the content. Use it where ambiguity is the enemy.