The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted user input (blocked or degraded prompts) to generate a revised version, creating an indirect prompt injection surface.
Ingestion points: The skill processes the user's original prompt text within the Repair Method logic.
Boundary markers: The skill contains explicit 'Boundary Rule' and 'Boundary — read before anything else' sections that instruct the agent to refuse unsafe requests involving minors, illegal material, or graphic content.
Capability inventory: The skill is limited to text generation and does not define or request access to tools for file writing, network operations, or command execution.
Sanitization: The instructions include a specific 5-step repair method designed to identify and replace risky surface wording with professional, non-graphic production language.
[EXTERNAL_DOWNLOADS]: The skill references external support files for its operation.
Evidence: References to ../../references/directors-read.md, ../../references/filter-vocab.md, and ../../references/multilingual-community-examples.md are included in the instructions.
Context: These are local file references within the vendor's repository structure (emily2040) used to provide context and vocabulary for the prompt repair task.

seedance-filter