The Agent Skills Directory

[COMMAND_EXECUTION]: The Rust implementation documentation demonstrates patterns for spawning and managing agent processes using tokio::process::Command. This functionality is essential for the orchestrator-agent architecture described in the skill.
[PROMPT_INJECTION]: The skill defines a communication surface that ingests external data from subprocess streams, which constitutes an indirect prompt injection surface. 1. Ingestion points: Data is ingested through sys.stdin in Python/Swift and the agent's stdout in the Rust orchestrator. 2. Boundary markers: The protocol utilizes line-delimited JSON as a structural boundary. 3. Capability inventory: The Rust component implements process management via command execution APIs. 4. Sanitization: The skill provides template code for message validation, including constraints on content length and required field verification.

ipc-communication