The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes transcript data which is typically provided by external users or files. This creates a surface where instructions embedded within a transcript could potentially attempt to influence the agent's behavior.
Ingestion points: Timeline transcripts processed in the SKILL.md instructions.
Boundary markers: The skill does not explicitly define delimiters or 'ignore' instructions for the transcript content, although it uses a template placeholder for the final output.
Capability inventory: The skill references cutagent-captions and cutagent-fusion CLI tools for caption generation and Fusion rendering.
Sanitization: No specific sanitization or escaping of the transcript text is mentioned before it is processed or inserted into templates.
[DYNAMIC_EXECUTION]: The skill uses a .setting file (Fusion configuration) as a template. The agent is instructed to dynamically modify this template at runtime by replacing placeholders and adjusting style parameters.
Evidence: The file assets/text-plus-caption-clean.setting contains a {{TEXT_PLACEHOLDER}} and comments directing the agent to perform 'Safe edits' on fields like Font, Size, and Color.

cutagent-transcript-captions