UI Microcopy & Labeling Audit

# Goal: UI Microcopy & Labeling Audit

You are working inside this repo. Mission: audit the small words that run the interface — button labels, field labels, tooltips, placeholders, helper text, confirmations, toasts — and find the ones that are vague, jargon-y, or leave the user guessing what happens next. This is the product's own words, not its marketing voice (17 covers that).

Read-only pass. Read the components, templates, and strings the UI renders, and run the app to see them in place; your only write is the report file.

## Phase 1 — Gather the strings
- Collect the interface microcopy: primary and secondary button labels, form labels and help, placeholders, tooltips, confirmation dialogs, toasts, section headings, nav items.
- Note where the strings live (components, i18n files, hardcoded) and whether they are centralized or scattered.
- Flag the highest-stakes words: the labels on destructive, paid, or irreversible actions.

## Phase 2 — Audit through 7 lenses
Cite the string and its location for every finding.
1. **Says what happens** — buttons name the action and its result ("Delete 3 files"), not "OK", "Submit", or "Yes"; the label that hides the outcome
2. **Plain language** — user words, not system or org jargon; the acronym, the internal codename, the database column surfaced as a label
3. **Placeholder abuse** — placeholders standing in for labels (gone the moment you type), or holding instructions the user needs while filling the field
4. **Helper where it helps** — the hint sits before the mistake, not only in the error after; format and requirements stated up front
5. **Consistent terms** — one name per concept across the UI; "delete" here, "remove" there, "archive" elsewhere for the same act
6. **Tone & length** — concise, calm, and human within the space; the paragraph on a button, the cute label that obscures
7. **Honest confirmation** — dialogs and toasts that name what happened or will happen, with an undo where truth allows, instead of a bare "Success"

## Phase 3 — Curate
- Rank by stakes × frequency: a vague label on a destructive or high-traffic action outranks a stiff tooltip seen once.
- Separate "unclear", "jargon", "inconsistent term", and "placeholder-as-label".
- For each, name the string, where it lives, and the rewrite.

## Phase 4 — Report
Create `MICROCOPY.md` at repo root:
1. **String findings** — each: lens · location · current text · the rewrite
2. **Term glossary** — the one word this product should use per concept, with the variants to retire
3. **High-stakes labels** — the destructive and irreversible actions to reword first
4. **The voice rules** — the handful of microcopy conventions (verbs on buttons, format hints up front) this codebase should adopt

Start the report with today's date. If `MICROCOPY.md` already exists from a previous run, read it first and lead with what changed since.

## Rules
- Every finding quotes the current string and gives the exact rewrite
- A button should name its action; if the label could be "OK", it is not done
- No user-facing interface copy in this repo? Say so in a one-paragraph null report and stop — a null result is a valid finding.
- If a `reports/` directory exists at the repo root, write the report there instead of the root.
- Before asking, present the top findings as a ranked list in plain words
- Report only — end by asking which strings to rewrite first