The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses the sensitive macOS Messages database located at ~/Library/Messages/chat.db to read message history and metadata. It also facilitates reading local file system paths to upload and send attachments.
[COMMAND_EXECUTION]: The local iMessage kit uses AppleScript (osascript) to programmatically control the Messages app for sending texts and attachments.
[PRIVILEGE_ESCALATION]: To operate locally, the skill instructs users to grant 'Full Disk Access' permissions to the process (terminal or IDE) within macOS System Settings, allowing it to bypass standard file system sandbox restrictions for the Messages database.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from incoming iMessages, which could contain malicious instructions.
Ingestion points: Inbound message handlers in references/local-v3.md (onIncomingMessage, onDirectMessage) and references/hosted-v2.md (subscribeEvents).
Boundary markers: The SKILL.md file contains a 'Common invariants' section explicitly warning developers to treat message text, attachments, and URLs as untrusted data and to avoid interpolating them into system prompts or commands.
Capability inventory: The skill possesses capabilities for reading local files, performing network uploads to hosted APIs, and sending automated replies.
Sanitization: While the documentation mandates sanitization as a best practice, the provided code snippets do not implement automated filtering or escaping of incoming text.
[EXTERNAL_DOWNLOADS]: The skill references package manifests and documentation from vendor-controlled domains and well-known services.
Fetches package manifests from unpkg.com.
Links to documentation and resources on photon.codes and GitHub repositories under the photon-hq organization.

imessage