The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill configures a messaging gateway that ingests and processes data from an external platform, creating a surface for indirect prompt injection.
Ingestion points: The gateway reads messages and commands from an external Buzz relay via the configured relay_url (SKILL.md).
Boundary markers: The skill advocates for security controls including require_mention: true to limit processing to targeted messages and allow_all_users: false combined with an allowed_users list to restrict the agent's interaction to authorized identities (SKILL.md).
Capability inventory: The agent possesses capabilities to execute shell commands via the buzz CLI, manage systemd user units, and modify local configuration and environment files (SKILL.md).
Sanitization: The instructions do not specify explicit sanitization or escaping of message content, relying instead on the underlying implementation of the Hermes gateway and Buzz plugin.
[COMMAND_EXECUTION]: The skill provides instructions for building and installing the Buzz command-line interface from source.
Evidence: Instructions include cargo build --release --locked -p buzz-cli and install -m 755 target/release/buzz /usr/local/bin/buzz to compile and deploy the binary to a system path (SKILL.md).
[PERSISTENCE]: The skill utilizes standard service management for maintaining the agent process.
Evidence: Instructions describe creating a systemd user unit drop-in at ~/.config/systemd/user/hermes-gateway.service.d/buzz-env.conf to ensure the gateway process loads environment variables and persists across sessions (SKILL.md).

buzz-hermes-gateway