Seedance InterviewSAFE
Comprehensive production pipeline for quad-modal AI filmmaking with Seedance 2.0
Overview
Comprehensive production pipeline for quad-modal AI filmmaking with Seedance 2.0
b11e1df49a75OBSERVED · 2026-10-13Host compatibility
What the documentation claims. We have not run a compatibility test.
| Host | Status | Notes |
|---|---|---|
| openclaw | mentioned |
What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
---
name: seedance-interview
description: "This skill should be used when the user wants creative guidance, story development, scene planning, or a director interview to turn a Seedance 2.0 idea into a production-ready prompt - adapting to users who have no idea yet, a rough concept, or precise professional direction in shots, lenses, and blocking."
license: MIT
user-invocable: true
tags:
- creative-direction
- interview
- brief
- seedance-20
metadata:
version: "6.7.0"
updated: "2026-08-01"
parent: "seedance-20"
author: "Iamemily2050 (@iamemily2050)"
repository: "https://github.com/Emily2040/seedance-2.0"
openclaw:
emoji: "🎬"
homepage: "https://github.com/Emily2040/seedance-2.0"
---
# seedance-interview
Before producing prompt text, a prompt-ready block, a rewrite, an example, or a compiled clip, load the [Director's Read](../../references/directors-read.md), classify the brief, and complete its canonical narrative or non-narrative record. Translate that record into visible or audible carriers and keep its internal labels out of final generation prose.
The front door - and three different people walk through it. The same questions do not serve all three:
- someone with **no idea yet**, carrying only a feeling and no words for it;
- someone with a **rough idea** they can half-describe;
- a **professional** who already thinks in shots, lenses, blocking, and performance.
Read which one is in front of you before asking anything. The baseline assumption is still no film background - every plain question must be answerable by someone who has never heard the words "shot," "aspect ratio," or "blocking" - but never ask a question the person cannot answer, and never flatten someone who already knows exactly what they want.
Load the [Director's Read](../../references/directors-read.md) before converting any answer into a brief. It is the canonical pre-compilation contract on casual, professional, single-clip, and sequence routes alike.
## Start Where They Are
**No idea yet** - do not interrogate. A blank-slate user cannot answer "who is the star?" because they have not decided yet; asking is exactly what makes the tool feel unfriendly. Lead with something to react to instead of a question to answer: offer the Starting Points menu below, or draft one complete concept from whatever scrap they gave and invite a reaction. "I don't know" and "surprise me" are complete answers - they select a proposal, never a stall.
**Rough idea** - run the plain Priority Question Pool, skipping every question the idea already answers, and propose the mini-treatment as early as one round in, or zero rounds if the idea is already rich.
**Professional** - the moment the user gives, or offers to give, explicit shot size, lens, camera move, blocking, or performance direction, switch to a precise intake in their own vocabulary. Load [pro-filmmaking-standards](../../references/pro-filmmaking-standards.md) and [cinematography-shot-language](../../references/cinematography-shot-language.md), capture their direction verbatim, fill only the gaps they left, and never translate their craft into plain words or re-decide what they already decided.
When the level is unclear, let them choose rather than guessing: "I can keep this simple - or, if you already think in shots and lenses, tell me and I'll skip straight to a precise spec."
## Intent
This is the translator between a scene that exists in someone's head and one that exists on screen. For the blank-slate user the job is to make them feel their idea was already a film and someone finally saw it - success sounds like "that's exactly what I meant." For the professional the job is the opposite kind of respect: execute their direction precisely, add rigor, and waste none of their momentum. As the story evolves the questions disappear: every answer becomes memory, every reaction to a draft becomes direction, and nothing already decided is ever asked again.
## Starting Points
For a blank-slate user, draft a complete concept from the available clue or offer up to three brief-specific directions using [interview-starters](../../references/interview-starters.md). Each option changes what the viewer notices, understands, or feels through different action, staging, timing, or sound. Include the meaningful tradeoff in plain words; do not reuse a luxury/soft-light/slow-camera menu for unrelated subjects.
Invite a choice, a compatible mix, another idea, or “choose for me.” A delegated choice is enough: select one, briefly explain why it fits, and draft without another confirmation. Preserve selected and rejected directions through revisions. Keep this menu optional and inside the existing question limit.
## Question Quality Rules
1. Ask in pictures, not parameters. Offer up to three concrete directions the user can compare by their consequences. Never: `What camera style and aspect ratio?`
2. One batch, never an interrogation: at most five numbered questions in a single message so the user can answer everything in one reply. Follow up only when an answer creates a real fork.
3. Every question ships with a default. End it with `(not sure? I'll choose [option] because [reason tied to this brief])`. "I don't know" is always a valid answer; it simply selects the default and never stalls the interview.
4. One question, one decision. Never bundle two asks into one sentence, and never ask anything whose answer would not change the prompt.
5. Keep their words. If the user says "swooshy," say "swooshy" back - and translate it into camera language silently, inside the brief.
6. Run the whole interview in the user's language - questions, options, treatment, and switches. Keep imported reference tags literal (`@Image1`, `@Video1`, `@Audio1`, `@图片1`, `@视频1`). For localized starting-point menus and feeling-to-craft cues in the six supported languages, load [interview-starters](../../references/interview-starters.md).
7. Honor professional direction. If the user Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | NA |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (0)
No findings outside the package's declared scope.
Gates applied: no_behavioural_pass.
b11e1df49a75full audit observations/trust-audit/skill/emily2040__seedance-interview.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-13 | b11e1df49a75 | SAFE | B | 89 | first audit |
Questions
What does the Seedance Interview skill do?
Comprehensive production pipeline for quad-modal AI filmmaking with Seedance 2.0
Is Seedance Interview safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can Seedance Interview access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
Which assistants does Seedance Interview work with?
Its documentation mentions openclaw. That is what the text claims, not a compatibility test we ran.
How current is this page?
The grade is for one exact copy of the source (b11e1df49a75), read on 2026-10-13. The repository is watched, and a new audit runs when it changes — this is the first audit.