Building a Shot List and B-Roll Prompt You Can Reuse on Every Project

Most of us rewrite the same prompt every time we sit down with a chat assistant, and every time it comes out slightly different, so the output comes out slightly different, and there is no way to tell whether the tool got better or worse. For a recurring task like generating shot lists and B-roll suggestions, that is backwards. The prompt is an asset. You build it once, tune it across a few projects, and reuse it — the same way you reuse a color LUT or an export preset.

This piece is about building that reusable prompt for shot lists and B-roll, and about the discipline that keeps it useful instead of generic.

Why a saved prompt beats a fresh one

A one-off prompt gives you a one-off answer. A saved, structured prompt does three things a fresh one can't. It enforces a consistent output format, so the shot list drops straight into your existing template instead of being reformatted by hand. It carries your accumulated corrections — every "stop suggesting drone shots we can't afford" you add becomes permanent. And it makes the tool auditable: when the output is bad, you can look at the prompt and see why, instead of shrugging.

In 2026 most assistants let you save this as a reusable object — a Claude Project, a custom GPT, or a saved system prompt in whatever your shop standardizes on. The mechanics differ; the principle is the same. Write it as a role plus rules plus a fixed output schema, and feed the variable material (the script, the constraints) at runtime.

The anatomy of the prompt

Here is the skeleton I use. I'll break down why each block earns its place.

ROLE
You are assisting a shooting producer on unscripted/documentary
work. You generate shot lists and B-roll suggestions from a script,
treatment, or scene outline.

HARD CONSTRAINTS (never violate)
- Only suggest shots achievable with the gear and crew I specify.
- Never invent locations, talent, or events not in the source.
- Flag any suggested shot that needs a permit, a release, or
  special access.
- Keep B-roll suggestions concrete and shootable, not abstract
  ("hands plating the dish," not "the essence of craft").

OUTPUT FORMAT (always this table)
| Scene | Shot # | Framing | Movement | Description | B-roll / cutaways | Notes |

For B-roll, prioritize shots that will cover edits (matching action,
transitions, and anything that hides a jump cut in an interview).

BEFORE YOU START
Ask me for: gear package, crew size, location(s), and total shoot
hours available, if I have not provided them.

The hard constraints block is the part that took me three projects to get right, and it is the part that makes the tool worth using. Left unconstrained, an assistant will cheerfully suggest a crane shot and a second unit for a two-person doc crew with a day rate. That is not a shot list, it is a wish. The "flag permits and releases" line has genuinely caught things a tired producer missed.

The output format block is what makes it reusable. Every project gets the same seven columns, so my template ingests it without reformatting, and my DP knows exactly where to look. The "cover edits" instruction is the single most useful line for post — B-roll's real job in unscripted is hiding cuts in interview, and an assistant that knows that suggests cutaways you will actually use instead of pretty filler.

Feeding it the right context

The prompt is only half of it. The other half is what you hand it at runtime. Vague input produces vague shot lists. I feed it:

• The actual script, treatment, or scene outline — not a summary. The assistant should work from the same document your DP is reading.
• Hard production constraints — gear, crew, hours, budget tier. "One shooter, one camera, FX3, no gimbal, four hours" produces a wildly different and far more useful list than the same script with no limits.
• The edit's needs — if you already know a scene will be a montage or will lean on a lot of interview, say so. The B-roll suggestions change completely.

A trick that pays off: after the first draft, I ask one follow-up — "which three of these B-roll shots are most likely to save an edit, and which three are the most skippable if we run short on time?" That triage is exactly the judgment a producer makes on the day when the light is going, and getting it on paper beforehand means nobody wastes twenty minutes on a shot that was never going to cut in.

Tuning it across projects

The prompt gets better only if you feed your corrections back into it. After each project I do a five-minute retro on the output. Did it suggest things we couldn't shoot? Add a constraint. Did it miss a category of cutaway we always need — establishing exteriors, tech inserts, hands-doing-the-thing? Add a reminder. Over a season, the prompt accumulates the institutional knowledge that usually lives only in a veteran producer's head, and it stops making the beginner mistakes.

Keep a changelog. When you tweak the constraints, note the date and why. Six months in you will want to know whether a change actually improved things or just changed them, and without notes you are guessing.

The honest limits

A few things this will not do, so you don't oversell it to your director.

It does not know your location. It has never seen the room, so it cannot tell you that the only clean angle is blocked by a window blowing out the exposure. Its framings are generic until a human who has scouted the space edits them. It also has no eye — it cannot tell you a shot is beautiful or dead, only that it is on the list. And it will confidently suggest coverage for a moment that, in the actual edit, wants to sit on a single unbroken take. Coverage is not always the answer, and the assistant defaults to more of it.

Treat the output as a thorough, consistent first draft that no shot gets missed from — a checklist that a human turns into a plan. The value is not creativity. It is that you never again walk onto a set having forgotten to list the establishing shot, the insert, or the reverse, because the prompt asks for them every single time. That reliability, applied across every project, is worth more than the occasional clever suggestion.

Build the prompt once. Correct it in public where your team can see the rules. Feed it real constraints. It will not replace a good shooting producer, but it will make an average shot list disappear as a task and free that hour for the things only a person standing in the room can decide.