Turn any idea, plan, or codebase into a beautiful interactive diagram. An agent skill for Claude Code, Codex, and more.
HTML agent-skillsai-agentsarchitecture-diagramclaude-codeclaude-skillscodexcoding-agentsdeepseek-harnessdeveloper-toolsdiagramsdiagrams-as-codedsh-pluginflowchartllmmermaidopencodesequence-diagramsoftware-architecturesystem-designvisualization
Ask explainer drafts to cover mechanism, checks and exceptions
tt-a1ia9919e826d15
diff --git a/archify/references/explainer-authoring.md b/archify/references/explainer-authoring.mdindex 794ff982..8509152c 100644--- a/archify/references/explainer-authoring.md+++ b/archify/references/explainer-authoring.md@@ -81,13 +81,14 @@ A diagram replaces the paragraph that would describe it. When the answer involve| A change to code | `@diff` or a ```` ```diff ```` fence || A command, config or code that does not exist yet | Plain fence |-A small answer gets a small page: the conclusion plus one or two panels.+A small answer gets a small page: the conclusion plus one or two panels. A how-it-works, diagnosis or checklist question usually needs three to five.## WritingEvery sentence gives the reader one fact they need. Write in the user's language.- Start with the fact. The title asks the question, the conclusion answers it, and the page ends when the facts end.+- Cover what the reader needs to act on the answer: the mechanism behind the conclusion, how to check it, and the cases where it does not hold. The next question a careful reader would ask gets a panel, a table row or a line in scope; a fact the reader could get wrong without it is not optional.- One sentence carries one fact; one panel answers one question. Say each fact once: when a table or diagram shows it, the text adds only what the figure cannot.- Call each thing by one name throughout the page, and use the name the reader already knows.- Give results with their evidence: a number with its unit, a file position, a source. When a value is unknown or estimated, say so in scope.