hey-tunelisted

# Tuning estimates **The user owns the final number.** This skill gathers the evidence, applies the decision to the ledger, and records why it changed. ```bash ROOT="${CLAUDE_PLUGIN_ROOT:-$PLUGIN_ROOT}" # Codex names it PLUGIN_ROOT HEY="$ROOT/scripts/hey.py" BOARD="$ROOT/scripts/board.py" ``` ## 1. Show the measurements first Never tune on instinct. Look at how finished items compared to their estimates. ```bash python3 "$HEY" variance python3 "$HEY" carryover --days 3 ``` `variance` measures business days from an item first starting to finishing, against its estimate. ``` P0|codegen pipeline est AI 0.4 -> 3 business day(s) open P11|locale resolution est AI 0.6 -> 4 business day(s) open ``` **These are elapsed days, not effort, and there is no mean to apply.** An item that was open three days may have had four hours of work in it: it waited for review, shared the days with other items, paused on a blocker. Walk them one at a time and **ask the user what share of each span went to the item.** That answer is the evidence; the span is only the prompt for asking. ## 2. Classify what changed Every tuning request is one of three things. **Decide which before touching anything** — they are handled differently. | Kind | What to change | |---|---| | **Scope grew** | Raise the estimate, and spell out the added scope as subitems | | **Estimate was wrong** | Change only the number, and note which multiplier was misjudged | | **Work did not finish** | Leave the es