The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to access and export sensitive user data. It targets the private application container located at ~/Library/Group Containers/group.com.apple.VoiceMemos.shared/ and directly queries the CloudRecordings.db SQLite database. This allows for the extraction of personal audio recordings and their associated metadata.
[COMMAND_EXECUTION]: The skill performs extensive command execution using osascript to run AppleScript and doShellScript within JXA (JavaScript for Automation). These commands are used to interact with the operating system, manage files, and execute external binaries like sqlite3 and python3.
[PRIVILEGE_ESCALATION]: The skill requires the user to grant high-level macOS permissions, namely Accessibility and Full Disk Access, to the automation environment (e.g., Terminal or Python). The setup instructions guide users through these steps, which grants the agent broad access to the filesystem and the ability to control other applications through UI scripting.
[DYNAMIC_EXECUTION]: The skill's documentation provides examples of dynamic code assembly. It uses JXA's doShellScript to execute shell commands constructed by concatenating strings with variable content, including transcript text passed to a Python subprocess for JSON processing. This pattern of runtime code assembly introduces risks if the interpolated content is attacker-controlled.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted content from voice recording transcripts.
Ingestion points: scripts/copy_transcript_to_file.applescript retrieves text directly from the system clipboard after performing UI actions in Voice Memos.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are used when processing the transcript text.
Capability inventory: The skill utilizes the Bash tool and shell execution via doShellScript and osascript.
Sanitization: There is no significant sanitization or validation of the transcript content before it is processed or written to disk.

automating-voice-memos