Demo runs — what is real here and what is not

Four complete runs of script-from-intake and claim-guard against one demo brand. The illustrations in ../../templates/ are drawn from these.

Read this page before the runs. Module 4's standard is that we show real output with honest provenance, and honest provenance here requires being precise about which parts of these runs are real and which are constructed, because they are not the same and the difference matters.

1What is real

• The output is genuinely generated. Every script in these runs was produced by following script-from-intake/SKILL.md against the inputs shown, by claude-opus-5, on 2026-08-14. It was not written by hand and then labelled as generated.
• The prompts are the actual prompts. Each run shows the brief exactly as it was filled from the intake, not a tidied-up version.
• The claim-guard findings are real findings. Nothing was planted. Where a run shows a rejected hook or an unsupported claim, the generator produced that line and the check caught it. Run 01 and Run 02 both ship a second choice because the first failed.
• The human edits are recorded, including the ones that made the output worse before they made it better.

2What is constructed

• Sable Kitchenware is not a real company. The brand, its guardrails and its contract terms are invented for demonstration. They are consistent with the BV-01 intake shown in mock 4, which is where they come from.
• Rae Okonkwo is not a real creator, and the experience file is invented.
• Therefore no result in these runs is evidence that anything works. They demonstrate that the skills behave as described. They do not demonstrate that a brand was happy, that a piece performed, or that a creator got paid.

3Why that distinction is doing real work

Module 4 draws a line between two claims: this is what we made and this is what happened when someone used it. These runs support the first and say nothing about the second.

That gap is why worked case studies remain an open item in the §3.1 asset inventory. A case study needs a real client, a real brief, a real approval round and a real outcome. Real output from invented inputs is a genuine improvement on hand-written illustrations — it is not a substitute for a case study, and this guide is not going to pretend otherwise about its own materials.

4The runs

_inputs.md holds the intake and experience file all four runs read from.