Wait — I don't understand where you've got to here. Re-pitch that: give me a
little bit of context, talk in ASD-STE100 Simplified Technical English, and use
this project's own nouns — product-context.yaml and docs/glossary.md where
they exist.
The register (for orchestrator skills)
The repair line above is also AEP's standing explain-to-a-human register.
Main-session skills that render status, escalations, or decisions for a person
(/aep-autopilot status and escalations, /aep-dispatch handoff summaries,
/aep-wrap layer-advance asks) write in it by default:
- One line of context first — what was happening when this became worth saying.
- ASD-STE100 Simplified Technical English — short sentences, one clause, active voice, one meaning per word.
- The project's own nouns — the ubiquitous language from
product-context.yamland the glossary, in place of invented terms and stacked acronyms.
A re-pitch is working if it is shorter and clearer, not shorter and blunter: it adds the premise that was missing instead of only deleting words, and it survives being invoked twice in a row without degrading into terseness.
This skill is user-typed only (/aep-easy-explain), and that is the design:
only the human knows when they stopped following. The model reaching for it on
its own would grade its own clarity — the same self-evaluation /aep-gen-eval
exists to forbid. Each host enforces it with its own native metadata: Claude
Code via disable-model-invocation: true in the frontmatter, Codex via
agents/openai.yaml (policy.allow_implicit_invocation: false); explicit
invocation stays available on both.
Adapted from mattpocock/skills
productivity/wait-what— the three-line original, whose insight is that naming the listener's state ("wait, you lost me") repairs comprehension, where naming the output ("be concise") produces telegrams.