Seedance 2.0 Shot Design

What to check before installing: - Review scripts/validate_prompt.py yourself (or have it reviewed) to confirm it contains no network calls, subprocess.exec usage, or unexpected file-system writes; if the script is run by the agent it increases risk. The SKILL README is inconsistent about whether the agent executes that script—clarify your agent runtime's behavior (does it permit local code execution?) before trusting automatic validation. - If you are concerned about accidental activation, note the skill's changelog says trigger phrases were greatly expanded; consider disabling autonomous invocation or restricting activation/triggers in your agent settings. - Because the skill reads local reference files, verify those files do not include secrets or references to external endpoints. The repository appears to use only documentation and templates, which is normal for this purpose. - If you want to be extra cautious: run the Python validator locally on sample prompts to inspect its behavior and outputs; if the agent environment forbids code execution, ensure the SKILL.md expects LLM-native validation rather than code execution. If you want, I can scan the validate_prompt.py file for network calls, subprocess usage, file writes, or obfuscated code and summarise what it does.