The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation recommends executing commands with elevated privileges to modify system binaries.\n
Evidence: The file references/installation.md instructs the user or agent to run sudo ln -sf "/Applications/Diffusion Studio.app/Contents/Resources/cli/bin/dapi" /usr/local/bin/dapi.\n- [EXTERNAL_DOWNLOADS]: The skill relies on downloading software and source code from external repositories that are managed by the vendor.\n
Evidence: references/installation.md includes instructions for brew install --cask diffusionstudio/tap/editor and git clone https://github.com/diffusionstudio/editor.git.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes and interprets content from external media files, which could contain malicious instructions designed to influence the agent's behavior.\n
Ingestion points: The dapi media listen and dapi media transcribe commands in SKILL.md ingest data from user-provided video and audio files.\n
Boundary markers: Absent; there are no instructions to the agent to treat media-derived content as untrusted or to ignore embedded instructions.\n
Capability inventory: The agent has access to the dapi CLI which can perform file system operations and media analysis.\n
Sanitization: No sanitization or filtering of the transcription or audio analysis output is performed before processing.

watch