Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add cogni-work/insight-wave/plugin install cogni-publishingWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cogni-work/insight-wave/design-verify)<a href="https://agentmods.dev/skills/cogni-work/insight-wave/design-verify"><img src="https://agentmods.dev/badge/skills/cogni-work/insight-wave/design-verify/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cogni-work/insight-wave/design-verify"><img src="https://agentmods.dev/badge/skills/cogni-work/insight-wave/design-verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00109 | $0.02653 |
| Opus 5.5 | $0.00044 | $0.01061 |
| Sonnet 5.5 | $0.00022 | $0.00531 |
| Haiku 4.5 | $0.00011 | $0.00265 |
Grade A, and why
design-verify scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Design Verify
Verify what design-render produced before anyone relies on it. The renderer already refuses a page or deck that fails its own fidelity checks; this skill checks the delivered file again, independently, against the frozen inputs, and adds what a render-time check cannot hold: per-family content preservation, object-level editability, declared accessibility, the critical visual classes, a persisted visual review, and a repair loop with a finite budget. The scripts decide what can be decided deterministically. A model or a person judges the rest, and only through a review record that names what was checked.
Inputs
- The frozen inputs of the render: the normalized brief and the semantic composition it was rendered from, and the theme directory it was rendered with.
- The delivered artifact:
index.htmlordeck.pptx; editability is inventoried from the deck package. - Optionally: a visual review record, a
design-rendermeasurement report (browser-report.json), and for a proof its manifest and specimen index.
Workflow
- Run
verifyon each delivered output. It directly extracts facts from index.html or deck.pptx using verification-owned readers and predicates, deriving expectations from the normalized brief, composition and pattern library; it grades preservation, editability (pptx), accessibility and geometry independently of renderer checks. Readdata.findings; an empty list is the only passing state. - Capture every unit at full resolution for the review: a full-page capture of the page at 1280 CSS px cut on each unit's border rows (each unit section is drawn with a 1 px border, so find its top and bottom border rows in the capture and crop between them with a host image tool), and each slide of the deck at its 1280 × 720 px canvas. Record the capture tool, version, resolution and image digest with each entry. Capture with host tools, never a plugin script: a headless browser screenshot for a page and a LibreOffice PDF export rasterized per slide for a deck, with the commands in
${CLAUDE_PLUGIN_ROOT}/docs/design-verify-proof.md§Reproduce it. - Review each capture and the deck as a whole under the rules below, recording each observation as a finding with a criterion, a severity (
critical,major,minorornote), a code and a description. A critical finding names one of the critical classes. The entry shapecheck-reviewandverify --reviewexpect is in${CLAUDE_PLUGIN_ROOT}/references/design-verify.md§Review record. The committed${CLAUDE_PLUGIN_ROOT}/docs/design-verify-proof/codex-openai/boardroom/pptx/review-record.jsonis a worked record that passescheck-review; copy its entry shape, and never add a bare verdict key such asoverall,qualityorscore, which the check rejects at any depth. - Run
verify --review <review-record.json>so the review's critical and major findings join the report. For a platform proof, runcheck-proof --manifest <proof-manifest.json>: it checks each artifact's review coverage and digests against its own frozen inputs. - When verification fails, run
render-verifiedinstead of hand-editing anything: it tries the other variants of the failing unit's pattern within the budget and returns either a passing render or a bounded failure with the repair history. - For a proof, run
check-specimensandcheck-proof;check-proofrecomputes every recorded hash and re-reads each report, the review and the specimen index.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed e10736d5fdde
- 1mo ago First seen · 91 lines · 109 tokens per session scan A 379629e818c2
design-verify is a skill published in the GitHub repository cogni-work/insight-wave (13 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 109 tokens to every session and 2,653 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-16.
Other skills, from other repositories
kami
A design and typesetting guide for creating professional documents and product landing pages, such as resumes, white papers, portfolios, and slide decks.
solidworks-engineering-drawing
A SolidWorks skill for creating and reviewing engineering drawings, which are manufacturing documents showing views, measurements, holes, parts lists, and title-block information.
document-design
Creates print-ready HTML that exports to PDF. Use to make a proposal, report, one-pager, newsletter, slides, or flyer.
design-export-repair
Fixes broken decks/PDFs exported from Claude's "Design" feature (or similar AI deck generators) — cut-off or clipped text, wrong/substituted fonts, and corrupted .pptx package structure that shows up only once you export to PDF, not while looking at it in the app. Use this skill whenever the user uploads a zip, .pptx…
architectural-drawing-parser
Parse architectural drawings, floor plans, and building code compliance documents using Vision AI. Extracts building type, occupancy, floor areas, room layouts, dimensions, and code parameters. Use when: reading PDF floor plans, analyzing architectural drawings, extracting building data from images or scanned…
paperlab_book_typesetting_release
Final production review for PaperLab books across HTML, PDF, DOCX, and ODF. Use for responsive layout, page breaks, equation rendering, table overflow, caption style, running headers, figure placement, and release-candidate polish.