The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local setup and configuration scripts, including setup/channels/imessage-configure.sh, scripts/photon-setup.ts, and setup/lib/restart.sh. It also performs build and validation steps using pnpm run build and vitest.
[PRIVILEGE_ESCALATION]: The instructions guide the user to grant 'Full Disk Access' to the Node.js binary on macOS. This is a significant privilege escalation that allows access to sensitive files across the entire disk, including the user's chat history (chat.db). This permission is explicitly requested and is fundamental to the 'Local' backend's functionality.
[INDIRECT_PROMPT_INJECTION]: As a messaging channel, the skill ingests untrusted data from iMessage conversations, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Inbound messages retrieved from the local chat.db or the remote spectrum-ts gRPC stream.
Boundary markers: The skill does not implement specific boundary markers or 'ignore' instructions for the incoming message data.
Capability inventory: The skill environment includes command execution (via shell scripts), file system access (via Node), and network operations (via the Photon SDK).
Sanitization: There is no evidence of sanitization or filtering applied to incoming messages before they reach the agent context.

add-imessage