The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes the adobe_execute_script tool, which allows the agent to run arbitrary ExtendScript code within the Adobe Premiere Pro environment to perform tasks not covered by standard API commands.
[EXTERNAL_DOWNLOADS]: Several commands, including import_mogrt, music_bed, and import_captions, allow the agent to fetch and import media assets from remote URLs provided by the user or external services.
[COMMAND_EXECUTION]: The skill makes extensive use of adobe_command to perform file system and application-level operations, such as creating sequences, applying cuts, and exporting projects to the local disk.
[DATA_EXFILTRATION]: The create_upload tool is designed to move local video footage from the user's disk to a remote service for transcription and analysis, which represents a managed data egress point for potentially sensitive media files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data via video transcription services.
Ingestion points: Untrusted data enters the agent context through the transcribe_video and analyze_transcript tools, which convert speech to text for the agent to process.
Boundary markers: The provided instructions do not specify any boundary markers or instructions for the agent to ignore embedded commands within the transcripts.
Capability inventory: The agent has access to powerful capabilities including arbitrary script execution (adobe_execute_script), file uploads (create_upload), and application manipulation (adobe_command).
Sanitization: There is no evidence of sanitization or validation of the transcript content before it is used to determine edit points or logic.

mira-premiere