v0.2.0 — Prompts That Travel
Author
🗓️ July 2026
Your .aether/docs are already a compressed map of your project. This release puts them to work for the AI you use outside Aether: the new /prompt command turns a plain task into a precise, file-referencing prompt you can paste into Claude, Kiro, Cursor, or anything else — so it goes straight to the right files instead of re-exploring your whole codebase, at a fraction of the tokens.
✨ Features
- New
/promptcommand. Type/prompt add a --json flag to the export command(no quotes needed) and Aether writes an optimized prompt to.aether/prompts/. It references the exact files to touch, the conventions that apply, and clear acceptance criteria — all drawn from your knowledge base - Never overwrites a prompt. Every run writes a new file with a meaningful name; a name collision becomes
<name>-2.md,<name>-3.md, and so on — your past prompts are always kept - Staleness gate. Before generating,
/promptdiffs your code against the last/genesissnapshot. If the docs are out of date it tells you what changed, recommends/sync, and asks whether to continue anyway. The generated file also carries a review-before-you-use reminder
🏗️ Under the hood
- Reuses the sync machinery — the freshness check runs on the same fingerprint diff (
loadSnapshot+diffFingerprint) that powers/sync, so a single project scan feeds both the staleness check and the live directory tree handed to the optimizer - Selective, low-token context — only the docs index,
AI_CONTEXT.md, andfolder-structure.mdare sent (bounded), keeping the call cheap; the point is fewer tokens, not shipping the whole knowledge base - Resilient output parsing — the model names the output file via a
SLUG:line; if it's missing, Aether derives a slug from your request and still keeps the full prompt, so a generated prompt is never lost - Cancel with
ESC— the single optimizer call is abortable mid-run, consistent with/genesisand/sync