The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes references to external video content on Douyin (douyin.com) within references/SOURCE-LEDGER.md. These links are used as instructional references for production techniques. As Douyin is a well-known service and the links are presented as static references without automated code execution, this is a safe use of external resources.
[PROMPT_INJECTION]: The skill is designed to process user-supplied narrative inputs, which introduces a surface for indirect prompt injection.
Ingestion points: User-provided scripts, story outlines, and oral descriptions are ingested in SKILL.md Section 1.
Boundary markers: The instructions mandate marking supplementary facts as assumed and missing data as pending in SKILL.md Section 1 and 5, providing a mechanism to distinguish external data from core instructions.
Capability inventory: The skill's primary function is orchestrating production contracts and invoking sub-skills for characters and scenes. It does not possess capabilities for sensitive file access, shell command execution, or network exfiltration.
Sanitization: Input data is mapped into structured JSON templates found in references/control-contracts.md, providing a layer of data organization, though specific string sanitization routines are not defined.

ai-short-drama-production