The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The terminal provider is designed to download a helper binary (tuichat) from the official vendor repository at github.com/photon-hq/tuichat upon its first execution to provide a TUI interface.
[COMMAND_EXECUTION]: The skill documentation describes the terminal provider spawning a subprocess to execute the tuichat binary, which is used to render the chat interface locally.
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to ingest and process messages from external platforms (iMessage, WhatsApp, etc.), there is an inherent surface for indirect prompt injection.
Ingestion points: Incoming message streams are processed via app.messages in getting-started.md and messages.md.
Boundary markers: The provided code samples do not explicitly implement boundary markers, leaving this to the developer implementing the agent logic.
Capability inventory: The skill provides capabilities for sending messages (space.send), replying (message.reply), and accessing local files for attachments (content.md).
Sanitization: The skill provides architectural guidance in best-practices.md for handling message pipelines, but developers are responsible for sanitizing content before passing it to downstream LLMs.

spectrum