[mrprkr]

Summary

• WAO page (/stuntkit/wao) caught up with @stunt-double/wao 2.2: audit mode (audit(), window.waoAudit, auto: false), the dialogs repair, action names for icons, hidden rules, plain-data reports, which agents it helps, Puppeteer, CDP and Playwright MCP injection, the full API, what it does not do, and upgrading from 1.x.
• Measured section: the fixture shop with and without WAO (0 of 10 vs 10 of 10 checkout steps by role), a real before/after from wikipedia.org, a table of six live sites with DOM size, pass time, repairs and issues, and size/cost/network/rendering numbers (4.3 KB gzipped, under 20 ms on a 4,300-node page, screenshots unchanged).
• Meta descriptions: every page already had a title, description, canonical, Open Graph and Twitter tags and a prerendered 1200x630 card. Fourteen descriptions ran past 160 characters, so search snippets and the OG card (which clips at 150) cut mid-word. Seven hand-written ones are shortened; generated skill pages now use summary() in scripts/generate.mjs, with a test.

Test plan

• pnpm test, pnpm type-check, pnpm build
• Script over the built HTML: all 71 docs pages have title, description, canonical, og:title/description/url/type/image (absolute, prerendered, width, alt), twitter:card/image; no duplicate titles or descriptions; none over 160 characters

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
stuntdouble-docs	Ready	Preview	Oct 27, 2026 12:16am UTC

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.

Not checked: tests were not run; formal verification proved 0 of 1 changed function(s) (1 not verified).

Formal verification. PR-changed functions: 0/1 verified (0 proven, 0 may-equivalent, 0 distinguished) · 1 not verified (1 vacuous).

Not verified on this run: mcpPages (vacuous: never exercised).

---

Graphify review — findings

Adds audit mode to the Web Agent Optimiser docs: audit() and window.waoAudit() report what WAO would change without keeping any of it, alongside a fuller before/after snapshot, a section on which agents benefit, and a single combined init script that works for Playwright, Puppeteer, raw CDP and Playwright MCP. Tightens the frontmatter descriptions on the API, Concepts, Index, StuntKit, browser toolset and Localeizer pages into shorter one-line summaries. Teaches the generator's summary helper to reduce a description to its first sentence, which throws on a skill with no description and cuts early at abbreviations like "e.g.".

Worth a look

• summary() throws on a missing skill description — scripts/generate.mjs:164 · Escalate · medium
  • NOT verified (no proof, no reproducing execution) — needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 101 functions depend on the 101 functions this change touches.

Health — this change adds coupling hotspots:

• worse: mcpPages() — 2 callers, 11 callees

Verification — 101 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 101 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify mcpPages.

The verifier did not have enough to check mcpPages, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 23 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly ReferenceError: splitSections is not defined — names the real obstacle, not a sampling gap)

· 1 more finding(s) on lines outside this diff (see the check run).