audit-agents
FreeAudit or rewrite AGENTS.md so it holds only lasting principles. Run it only when the user asks for it by name; never invoke it on your own.
The English summary is being prepared. Showing the original description.
Shoot software UI like a film. A sequence of hero shots, each a steep, oblique, extreme close-up of the interface floating in black, with a razor-thin focal plane, a slow motivated camera drift, and UI that acts on its own, hard-cut on the action, ending on one line of copy and a still logo. Use when asked for a launch video, product film, feature trailer, teaser, promo clip, hero shot, or any "cinematic" footage of an app or UI. This skill covers camera, framing, light, and edit. It is stack-agnostic.
The English summary is being prepared. Showing the original description.
See how to installReview the instructions the agent will receive before you install.
You are the director and the DP, not the motion designer. The film is a sequence of hero shots. Each one looks at a real piece of the product's UI from a steep, low angle, so close that the focal plane covers only a line or two of text and everything else melts into black. The camera drifts slowly toward where the next thing will happen. The UI does the work by itself: things get typed, streamed, selected, and written while we watch. Nothing explains itself with captions.
This skill is about camera, framing, light, and the edit. Build the film with any rendering stack. Whatever you choose, make every frame a pure function of time, so the film renders deterministically and can take motion blur.
Read references/look.md for exact camera, depth-of-field, light, reveal, and type recipes. Read references/shot-list-example.md for a full example of the shot grammar.
For a ~15 second promo video made in code from a landing page (a continuous 3D fly-over of the page ending on a title card with the brand, logo mark, a two-part closing line, and an optional track), follow references/landing-page-promo.md.
A black void. The product floats in it like a dark slab, tilted away from the camera, lit by a soft pool that follows whatever matters and lets everything else fall into black. The camera is low and close, looking along the UI rather than at it, so lines of text run diagonally across the frame and shrink into the distance. The focal plane is a thin band: one control, one line of typing, one row of a list is sharp, and the rest is soft bokeh. The camera never stops drifting and never moves fast. Shots are long enough to watch something happen. The only colour is the product's own. Restraint everywhere else.
Before planning a single camera, write down:
Post the shot list before building, then keep going:
| # | Cut at | Len | Subject (what is in focus) | Scale | Angle | Move | Action |
|---|
Rules:
Measure scale as magnification of the UI at the focal point, relative to its native size:
| Scale | Magnification | Use for | Depth of field |
|---|---|---|---|
| Extreme macro | 5–7× | A single tab, a button, a nav item, a status badge. One or two words fill a third of the frame. | Only the subject is sharp. Everything else is heavy bokeh. |
| Macro | 2.5–4× | A composer with a line of typing, a few rows of a list, a code hunk | A band 2–4 text lines deep is sharp |
| Medium | 1.5–2.5× | A panel at work: a list streaming in, a document writing itself, a diff with a sidebar | A wider band is sharp; the far edge and near edge stay soft |
| Wide | 0.8–1.1× | The payoff only: the whole window, near-frontal, with content scrolling or changing | Mostly sharp; the frame edges still soften |
Macro shots must stay crisp. Rasterise the UI at the resolution of the magnification (for example 6×) before the camera scales it. Never upscale a 1× bitmap.
This is the signature of the look. Except for the wide shot, the camera is low and oblique, looking across the UI plane:
Each shot is one continuous move. It is motivated: the camera drifts toward where the next thing will appear.
Every shot must look like it was filmed on the same stage. Values are in references/look.md.
Add no CTA buttons, URL cards, or "available now" banners unless asked.
There is no music: no score, no drone, no beat, no risers, no braams, and no tonal stings. The soundtrack is quiet, tactile foley over near-silence.
Every visible action gets a cue: a soft key tick per typed character, a soft click when a tab or item is selected, a quiet whoosh on send, and a faint granular hiss under streaming text. The payoff gets no impact. The copy and the logo play over silence.
These make a film look like a tutorial or a template:
Render a still at the middle of every shot, at each action's peak, at the payoff, and at the copy and logo frames. Judge each one as a frame grab from a film:
Iterate until every still could be a screenshot in a launch post.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Audit or rewrite AGENTS.md so it holds only lasting principles. Run it only when the user asks for it by name; never invoke it on your own.
The English summary is being prepared. Showing the original description.
Audit the choices an implementing agent made, not its diff — a pure decision audit that traces the session's history into a choices ledger, changes no code, and never blocks an unsupervised run. Working code still embeds architecture the user never chose; surface it because future work inherits it. Use when the user wants to review the decisions the AI made on their behalf, before merging or committing AI-implemented work, when integrating a delegated subagent's pass, or when a fix "works" but might be a point fix.
The English summary is being prepared. Showing the original description.
Audit and prioritize performance work through bounded-work and forward-progress checks. Use when asked to find performance issues, investigate freezes/500s/OOMs, review retries or polling for no-progress loops, rank a performance backlog, or implement the next simple performance fixes.
The English summary is being prepared. Showing the original description.
Audit whether tests earn their maintenance cost and which suite owns each contract. Use when pruning redundant tests, investigating implementation coupling or test-only production hooks, reviewing the value of proposed coverage, or auditing an entire subsystem. Use write-tests to implement the resulting test changes.
The English summary is being prepared. Showing the original description.
Run fast, progressive experiments to map tunable parameters and their effects. Use when optimizing code, prompts, configurations, or other artifacts against a goal or benchmark, or when a long research run needs a planned hypothesis queue, focused trials, and durable findings.
The English summary is being prepared. Showing the original description.
Use Claude Code as an independent `claude -p` subagent when the user explicitly asks for Claude, wants a second-agent opinion from Claude, or asks to delegate a well-scoped task to Claude. Supports selecting `--model` and thinking/effort level with defaults of the current `opus` alias and `high`.
The English summary is being prepared. Showing the original description.