show-me-dont-tell-melisted

# show-me-dont-tell-me Every step = **one annotated image + one line of action**. The text never describes what the image already shows. Answer in the user's language. ## Absolute rule Never describe an interface from memory. Real capture or source code first, writing second. Interfaces change between versions; an invented instruction costs more time than it saves. If nothing is obtainable: say so and ask for a screenshot. Never guess where a button is. --- ## Step 0 — Triage the request Four modes. The mode decides everything downstream. If the request is ambiguous, ask **one** question to settle it. | Mode | Triggered by | Captures | Output | |---|---|---|---| | **LIVE** | "help me with", "I'm stuck", a session in progress | 1 per message, recapture before each instruction | short message + image | | **DOC** | "document this", "tutorial for my app" | 1 per step + edge states | structured document, images inline | | **TROUBLESHOOT** | a specific error, "it's not working" | capture of the error + of the fix | cause → fix → verification | | **SCRIPT** | "video", "screencast", "training" | 1 per shot | timed script, numbered shots | ### Audience Identify who it is for — it changes vocabulary and how much "why" to include: - **End customer** — zero technical jargon, never a server file path, reassuring tone on errors. - **Internal team** — house jargon allowed, keyboard shortcuts, edge cases. - **Your future self** — record decisions and reasons, not only gestures.