Reviewing a Builder artifact with Wally — itemized suggestions you apply one at a time

Wally Review is the same idea applied to every Builder surface: ask for a full editorial pass and get back a list of small, concrete changes — each with a before/after diff and a one-sentence rationale — that you accept or skip individually. Most of the time you'll keep some, drop some, and end up with copy that's noticeably tighter than what you started with.

Where it lives:
• Builder Docs — right rail, below the selection-rewrite panel. Always visible while a doc is open.
• Sites — floating "Review with Wally" button on the page editor. Click to open the review drawer.
• Slide Decks — floating "Review slide with Wally" button. Reviews the active slide's text objects.
• Design Studio — floating "Review with Wally" button. Reviews text objects on the active page.
• Socials Composer — floating "Review with Wally" button (bottom-left). Reviews the post copy + name.

How a review works (the same in every editor):
1. Open the artifact in its editor and (optionally) type a focus into the panel — e.g. "tighten the copy", "more donor-friendly tone", "make it scannable for federal reviewers". Empty focus is fine; Wally will use a default editorial rubric tuned to the artifact kind.
2. Click "Ask Wally to review". Wally reads the whole artifact and returns 3–8 itemized suggestions, sorted critical → improvement → polish.
3. Each suggestion shows:
• A severity dot (critical / improvement / polish) and category chip (copy, tone, layout, structure, a11y, seo)
• A short title — what the change is
• A one-sentence rationale — why it helps
• A word-level before/after diff (red strikethrough for what's removed, green highlight for what's added)
4. For each suggestion you have three buttons: Apply (replaces the original text in place), Skip (dismisses it), or — if you've already applied — Revert (puts it back).
5. Hit "Apply all" at the top to apply every pending suggestion in one click. You can still revert each one individually afterward.

What Wally is graded on, by artifact:
• Docs — clarity, tightness, active voice, evidence behind capability claims, doc-kind-specific rubric (Section L compliance for RFI responses, outcome-led prose for impact reports, etc.)
• Sites — punchy headlines (benefit before feature), tighter body copy, stronger CTAs (verb + outcome), scannability
• Decks — slide titles that make a single point, bullets that are short / parallel / progressive, no wall-of-text body copy
• Design Studio — punchy headlines, minimal body copy, captions that don't repeat the visual
• Socials — first-100-character hook, channel-fit, CTAs / links / hashtags that earn their place

Itemized = real choice:
The whole point of itemized review is that you're not stuck with "take Wally's whole rewrite or nothing". You'll usually like some suggestions and disagree with others — apply the ones you like, skip the ones you don't, and the artifact ends up better than before without losing your voice.

Limitations to know:
• Decks + Design Studio review the ACTIVE slide / page only. To review another slide or page, navigate to it and run again.
• Wally pins each suggestion to a specific text snippet that was present when you ran the review. If you've heavily edited that span between running and applying, the apply will fail with "Original text not found" — just run the review again on the new state.
• Tilted, deeply-formatted, or template-bound text is reviewed the same as any other text — the diff is over the literal characters.

Cost transparency:
Each review uses Anthropic Claude tokens. Token totals (input + output) roll up to your org's monthly usage at orgs/{orgId}/usage/{yearMonth}.aiTextTokens, billed on the standard +15% chargeback. Counter at the same path: orgs/{orgId}/usage/{yearMonth}.wallyBuilderReviews.