Questions from the script desk
Untangle the testing questions
Use these answers to set expectations before choosing a tool or starting a review pass.
Can AI certify that a branching story is consistent?
No. A model can suggest checks and candidate conflicts. Its output does not prove narrative correctness or complete path coverage. Even formal checks apply only to the rules and bounds encoded in the model.
What is the difference between state testing and dialogue review?
State testing checks recorded facts and transitions against rules. Dialogue review checks what a line asserts and whether the speaker and context support it. A correct state can select an inappropriate line; an appropriate line can conceal an incorrect state.
Does visiting every scene mean every branch was tested?
No. Scenes can contain conditional lines, different entry states and hidden guards. Report passage, edge and condition coverage separately. State the starting configurations and search limits so readers know what the result actually covers.
How should we handle false positives?
Keep the original evidence, check the missing context and record why the finding is not a defect. Scope exceptions narrowly. An approved lie in one scene is not a reason to suppress every future warning about the character.
Can a character deliberately contradict the game state?
Yes. Lies, mistakes and unreliable narration can be intentional. Review world truth, character belief and spoken claims separately. Ask the writer whether the effect is intended before trying to make every statement objectively accurate.
Do these methods work with ink, Yarn Spinner and Twine?
The principles apply, but extraction and runtime instrumentation depend on the tool, version, story format and host integration. This site supplies no universal adapter. Twine story formats in particular can differ in their variables and execution behaviour.
Can I upload a script to this website?
No. This website has no script upload or analysis service. The contact tool only prepares a local text note. Copying or downloading that note does not send it to us.
What belongs in a reproducible narrative ticket?
Include the build, starting state, exact choices, line identifiers, observed state, expected behaviour source and any unresolved interpretation. An optional screenshot or recording should support those details rather than replace them. See the reporting guide.
How should confidential scripts be handled with external AI tools?
Use approved services and review access controls, retention, training terms and contractual safeguards before sharing unpublished material. A consumer chat interface and an enterprise deployment may have different terms. Do not assume a tool is suitable merely because it can accept a long script.
Still an open question?
Send a specific editorial question or correction without including confidential game material.
PREPARE A NOTE