Case study media — template, art direction, storyboards
doingOne presentation template across all ten case studies, a single art direction, an image brief and a full video storyboard per study. Prompts are the deliverable; rendering comes after they read right.
Task graph
Ordered the way the dispatcher would work it. Each row shows the loop's own verdict —
this is the same computation scripts/next.mjs runs, not a second opinion.
- csm-002 The presentation template — what every case study shows doingneeds csm-001
5 acceptance criteria · 1 run
- Defines an ordered frame set with each frame marked mandatory or optional, and says what job each frame does
- Every frame specifies aspect ratio and whether the source is Figma export, live screenshot, or generated
- Explicitly handles the case of a study with no existing assets
- Validated against two real studies of different shapes — one with a Figma link (zevo) and one with none (agent4)
- A person could apply it to an eleventh case study without asking a question
This is the task foogy actually asked for: “I have all the files in Figma, I just need a template of what’s best to show.”
The template is a decision about what to pull out of a Figma file, not a layout. A case study that shows six screens shows nothing; the frames have to carry a narrative — the problem, the mechanism that solves it, the proof it worked.
Use the Figma MCP to look at real files where links exist (haas, hive3, traverse, zevo) rather than theorising about what is in them.
The validation criterion is the honest one: a template that only works on the study it was designed against is not a template. zevo has a Figma link and six assets; agent4 has neither. If it handles both, it holds.
Log
- 2026-07-26 — claimed, attempt 1/3.
- 2026-07-26 — verified pass. Parked for approval (lane 2).
- ok 2026-07-26 attempt 1 $1.80 Template: 5 mandatory frames + 2 optional, wired to existing slots. Validated against zevo (assets) and agent4 (none).
- csm-003 Art direction — one visual language across all ten todoneeds csm-002
5 acceptance criteria
- Names palette, lighting, camera treatment, composition rules and texture in terms a prompt can use
- Distinguishes ux-ui from web3-ai studies without breaking the shared system
- Sits with the existing light theme and #ff5a1f accent rather than fighting it
- Includes three anti-patterns — what would make these look like generic AI stock
- Expressed as a reusable prompt fragment that every later image and video prompt inherits verbatim
The marketing decision: how should these look. One direction across ten studies spanning 2020 to 2026, enterprise e-commerce to on-chain agents.
The reusable-fragment criterion is what makes this real rather than a mood board. If the direction cannot be written as text that drops into every prompt unchanged, it will not survive contact with ten separate briefs, and the set will drift into ten different looks.
The anti-patterns matter as much as the direction. Generated marketing imagery has a house style — glowing gradients, floating glass panels, teal-and-purple — and landing there would make a decade of real work look synthetic.
- csm-004 Image briefs + prompts — all ten studies todoneeds csm-003
6 acceptance criteria
- One brief per study, all ten, each following the csm-002 frame set
- Every frame carries a complete generation prompt: subject, treatment, aspect ratio, and the art-direction fragment verbatim
- Each prompt states what the frame does in the narrative, so a wrong render is diagnosable
- Frames sourced from Figma or a live screenshot are marked as such and get no prompt
- agent4 and depunks are fully covered despite having no starting assets
- Published as src/docs/foogy-xyz__media-brief-<slug>.md, one per study
Ten briefs. The criterion that stops these being decorative is naming what each frame does: when a render comes back wrong, “the frame was supposed to show the owner earning while the car sits idle” tells you what to fix. “Hero image” does not.
Marking Figma-sourced frames as no-prompt is equally important — the constraint is that real work outranks generated work, and a brief that quietly proposes generating a screen that already exists has broken it.
Runs in parallel with csm-005.
- csm-005 Video storyboards + full prompts — all ten studies todoneeds csm-003
7 acceptance criteria
- One storyboard per study: shot list with duration, camera move, subject, and transition
- Total runtime stated per study and defensible for where it will be used
- Every shot carries a complete, self-contained generation prompt — usable months later with no other context
- Names the first frame source for each shot: generated, Figma export, or live screen recording
- Opening shot earns attention in under two seconds; closing shot lands the outcome
- No shot depicts a product state that never existed
- Published as src/docs/foogy-xyz__storyboard-<slug>.md, one per study
This mission renders no video. The deliverable is the storyboard and the prompts, exactly as asked — video is the most expensive thing to get wrong, and a prompt that reads right on the page is the cheapest possible place to find out it doesn’t.
The self-contained criterion is the one that matters in three months: a prompt that only makes sense next to this document is a prompt that will be rewritten from scratch.
The last criterion is a constraint, not a style note. A shot showing Samsung’s webshop doing something it never did is a false claim about a real client.
Runs in parallel with csm-004.
- csm-006 Generate the image set — batched, approved, capped todoneeds csm-004
5 acceptance criteria
- One batch per study, submitted only after that study's brief is approved
- Every generated file traceable to the brief line that asked for it
- Running spend recorded per batch; stops at the mission cap without being told twice
- A rejected frame is re-prompted from the brief, not hand-tuned into something the brief never described
- Output lands in mfugaj/foogy.dev under src/assets/images/case-studies/<slug>/
The only task here that spends money.
Higgsfield calls require explicit approval in this environment — even the balance check is gated — so authorisation is enforced by the platform rather than by good intentions. Batch per study so approval is granted once per brief instead of once per frame.
The re-prompt criterion protects the brief’s authority. Hand-tuning a prompt until a frame looks nice produces images nobody can explain later and a brief that no longer describes what shipped. If the render is wrong, either the prompt was wrong — fix the brief — or the frame was wrong to ask for.
Note the different repo: briefs live in dash, media lands in foogy.dev.
- csm-007 The overview page — all ten, side by side todo
4 acceptance criteria
- One dash page showing all ten studies with their frame set and storyboard status
- Makes the set legible as a system — obvious at a glance which studies are complete and which are thin
- Links through to each brief and storyboard
- Readable at 375px with no horizontal scroll — mobile-audit stays clean
The marketing/design overview itself — the thing that makes ten separate documents into one view.
Its real job is exposing imbalance: if three studies have eight frames and seven have two, that should be visible immediately rather than discovered while building a portfolio review.
The mobile criterion is not boilerplate. dash-mobile exists because this dashboard was unusable on a phone; a new page that reintroduces overflow would undo it, and the audit gate will catch it.
- csm-001 Asset inventory — what exists, what is missing, per study done
4 acceptance criteria · 1 run
- A table covering all ten studies: existing image files, Figma link, live URL, category, year, and whether a real product screenshot exists
- Each existing asset classified as cover / UI screenshot / diagram / photo / decorative
- Names every gap explicitly, including the two studies with zero assets
- Published as src/docs/foogy-xyz__media-inventory.md so it renders on the dash
Before deciding what to show, establish what there is. The counts in the mission spec are a starting point, not the inventory — 5 files in a folder says nothing about whether they are usable UI captures or a stray logo.
Read the actual files in
mfugaj/foogy.devundersrc/assets/images/case-studies/, and the frontmatter of eachsrc/content/case-studies/*.md.Lane 1: it produces a document and changes nothing.
Log
- 2026-07-26 — claimed, attempt 1/3.
- 2026-07-26 — verified pass. Auto-merged (lane 1).
- ok 2026-07-26 attempt 1 $0.60 Inventory: schema already has gallery/diagram/video slots, all unused; 24 image files on disk unreferenced; agent4 and depunks have zero assets.
The spec
Case study media
Ten case studies on foogy.xyz. The work behind them is strong and the assets exist — they just aren’t presented as a system, and there is no motion at all.
What exists today
| Case studies | 10 — zevo, wire, samsung-eshop, agent4, depunks, haas, hive3, liberoverse, sancta-domenica, traverse |
| With image assets | 8 of 10 (5–6 files each). agent4 and depunks have none |
| With Figma links | 4 of 10 — haas, hive3, traverse, zevo |
| With motion | 0 |
| Categories | ux-ui and web3-ai |
Two gaps decide the shape of this mission: the two newest and most strategically important studies (agent4, depunks) have no imagery at all, and nothing anywhere moves.
Goal
A repeatable way to present any case study, applied to all ten, so the set reads as one body of work rather than ten separate efforts — plus a video storyboard per study, written as prompts precise enough to render later without rethinking the idea.
Scope
- A presentation template. What a case study shows, in what order, at what aspect ratios, and which frames are mandatory versus optional. The assets are already in Figma; the missing thing is a decision about what to pull out of them.
- One art direction across all ten — palette, lighting, composition, treatment — so a viewer can tell they came from the same studio.
- Per study: an image brief with a full generation prompt per frame.
- Per study: a video storyboard — shot list, timing, motion, and a complete prompt per shot.
- Then, and only then, generation.
Non-goals
- No rendered video in this mission. Storyboards and prompts are the deliverable. Video is the most expensive thing to get wrong and the most expensive to redo; the prompt has to read right before anything is spent.
- No rewriting the case study copy. This is presentation, not editorial.
- No portfolio redesign. The media has to fit the existing light theme and
#ff5a1faccent as it stands. - Not replacing the existing 8 covers wholesale. Where a real screenshot is better than a generated image, it stays — this is not an excuse to make everything synthetic.
Constraints
- Real work outranks generated work. These are case studies about things that actually shipped. Generated imagery frames, contextualises and adds motion; it never fabricates a product that did not exist or a screen that was never designed. Anything that could read as a claim about the client’s real product is out.
- Higgsfield generation calls require explicit approval in this environment — the balance check alone was gated. Treat every generation as human-authorised, and design the tasks so a batch is approved once rather than a prompt at a time.
- Figma is the source of truth for what was actually designed. Where a Figma link exists, the template must be checkable against it.
- Deliverables are markdown docs in
src/docs/foogy-xyz__*.mdso they render on the dashboard and can be read on a phone.
Acceptance (mission-level)
- One template document that a person could apply to an eleventh case study without asking a question.
- Ten image briefs, each with per-frame prompts that name the aspect ratio, subject, treatment, and what the frame is doing in the narrative.
- Ten video storyboards, each with a shot list and a complete prompt per shot.
- agent4 and depunks have a full plan despite starting with zero assets.
- Every generated frame is traceable to the brief line that asked for it.
The line
Prompt-writing, briefs and storyboards are lane 2 but cheap and reversible.
Generation is where money leaves. It is gated behind an approved brief, runs as one batch per study rather than ad hoc, and stops at the budget cap. No task in this mission renders video.