Vinson's locally-running HTML print checker returns wrong results on color profile and bleed compared to Acrobat, blocking team adoption and leaving manual Acrobat checks as the only reliable path. · Seen in: 2026-06-24-0800-ai-design-gang-dQyU.md

A fly.io worker that accepts a PDF upload, runs PyMuPDF or pdfinfo to extract color profile, bleed dimensions, DPI, PDF version, and artwork size, compares against a configurable spec sheet, and posts a pass/fail report with field-level detail to the existing Slack print-check channel.

Extend Vinson's existing print checker concept by replacing the browser-side PDF parsing with a fly.io worker running PyMuPDF. The worker exposes a single POST endpoint that accepts a PDF binary, extracts the five fields already in Vinson's spec, and returns a structured JSON result. A thin Slack slash command or file-upload listener triggers the worker and formats the result as a Slack message with green/red field indicators. Validation against 10 known-good and 10 known-bad files from the existing Slack print-check channel history serves as the accuracy gate before team rollout.

Every print file currently requires a manual Acrobat check, and a single missed bleed or wrong color profile triggers a costly reprint, so the cost of a wrong automated result is high enough that the tool cannot ship until it matches Acrobat on a documented test set.

URL-based branding extraction reliably pulls granular tokens but consistently fails to capture compositional brand character such as spatial rhythm and background texture, forcing multiple Claude revision loops before any on-brand output is usable. · Seen in: 2026-06-24-0800-ai-design-gang-dQyU.md

A structured Figma annotation layer where designers tag three to five reference frames per client with compositional labels such as dominant texture, spatial density, and accent behavior, which a Claude agent then reads via Figma MCP to generate a brand character spec that supplements the token extraction output.

Build a /brand-intake skill that reads a Figma file URL and a set of designer-annotated frame IDs via Figma MCP. Claude extracts the tagged frames, reads the annotation labels, and generates a brand character spec in JSON covering five compositional dimensions: spatial density, texture type, accent placement pattern, typographic contrast ratio, and motion tendency. This spec is stored alongside the token output in the client brand profile folder and is passed as context to any subsequent ad, wireframe, or illustration generation task. The first pilot client is one of the existing OTP accounts with a complete Figma file already in the system.

Branding extraction is Willi's stated prerequisite for the entire AI design workflow roadmap, and every week it remains stuck on the look-and-feel problem delays the wireframe, illustration, and ad automation layers that depend on it.

Subscription clients like Falkemedia cannot formulate Webflow-compatible feedback because they think in page-builder baukästen logic, causing tickets to stall mid-flight waiting on client input that never arrives in a usable form. · Seen in: 2026-06-24-1359-falkemedia-x-magier-8Efo.md, 2026-06-23-0800-otp-pulse-check-4ZGe.md

An n8n workflow triggered when a Many Requests ticket moves to a client-review status that checks whether the client organization has a Webflow literacy Loom assigned, and if not, automatically attaches the relevant Loom links from a curated library based on the ticket category tag such as image swap, variant switch, or background color.

Build a small Loom library of three to five recordings covering the actions Falkemedia and similar clients actually need: swap an image, switch a header variant, set a mobile background color, and navigate the component library. Store the Loom URLs in a markdown lookup table keyed by ticket category tag. An n8n flow triggered by the Many Requests ticket status webhook checks the client organization record for a literacy-onboarded flag, and if absent, appends the relevant Loom links to the ticket activity feed as a PM-visible note before the client review message goes out. The Falkemedia account is the first pilot since Nicolaus already committed to a one-hour intro session that can be recorded as the source material.

Tickets stalling on missing client feedback directly extend OTP project timelines and compress the design execution window, and the Falkemedia case shows the pattern repeats across clients who come from page-builder backgrounds.

Animation revision cycles escalate from minor zoom and transition tweaks into full re-renders because there is no structured scope gate between feedback receipt and the decision to re-render, costing overnight render time on work that was already close to done. · Seen in: 2026-06-23-0959-animation-nuwacom-N2fn.md

A Claude agent that reads a feedback comment thread from a ticket or Slack thread, classifies each feedback item as a compositing edit, a render-required change, or a direction reset, and posts a structured scope summary to the ticket before the animator begins work, flagging any item that would require a full re-render for explicit PM sign-off.

Build a /animation-scope-gate skill that accepts a Granola transcript basename or a Slack thread URL as input. Claude reads the feedback items, classifies each against three categories using a simple rubric: compositing edits are opacity, zoom, cursor, and transition changes that can be done in post; render-required changes are lighting, environment, camera angle, and background changes that require re-rendering; direction resets are concept-level changes that invalidate the current version. The output is a three-section markdown summary posted to the relevant ticket. Any render-required or direction-reset item requires a named PM to approve before the animator proceeds. The Nuwacom case is the first pilot since the feedback from June 17 was compositing-only but was interpreted as a direction reset, triggering an unnecessary overnight render.

The Nuwacom animation session shows that a single misclassified feedback item cost at least one overnight render cycle and a same-day emergency call to reset direction, and the pattern will recur on every animation project that involves iterative client feedback.

Subscription clients with both ongoing web design queues and urgent campaign asset requests have no structured way to signal dual-pipeline priority, forcing PMs to manually negotiate queue order in kickoff calls instead of having it encoded in the ticket system. · Seen in: 2026-06-23-0807-dexxter-website-rebrand-part2-kickoff-11R5.md

An n8n workflow that detects when a client organization has two active ticket streams tagged with different pipeline labels such as web and campaign, automatically creates a linked priority table in the ticket system with the client-supplied order, and posts a daily digest to the assigned PM showing the current top item in each pipeline and its age.

Build an n8n flow triggered by a new ticket creation event in Many Requests. If the creating organization already has an open ticket in a different pipeline category, the flow posts a prompt to the PM Slack channel asking whether a dual-pipeline split is needed and, if confirmed, creates a pinned priority table in the organization's ticket view with two columns. The daily digest runs as a scheduled n8n task each morning, reads the top open ticket in each pipeline per organization, and posts a two-line summary per dual-pipeline client to the PM channel. The Dexxter account is the first pilot since Leya and the Teamleader team explicitly negotiated the split in the kickoff call and the web dev agency dependency makes ticket order externally consequential.

Dexxter's July peak period starts within days and the web dev agency consumes designs in priority order, meaning a mismanaged queue directly delays the agency's implementation schedule and creates a client-visible bottleneck.