The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill guides the user to download a deployment script from a placeholder URL (https://example.com/deploy-resonate.sh) and execute it using sudo. This pattern of piping remote content to a shell with elevated privileges is a high-risk practice that can be exploited if the source or connection is compromised.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions to download a binary executable for jwt-cli from a third-party GitHub repository (github.com/mike-engel/jwt-cli). Fetching and running binaries from personal or unverified external repositories introduces supply chain security risks.
[COMMAND_EXECUTION]: The deployment process relies heavily on sudo for administrative tasks, including binary installation, directory permission changes, and systemd service management. This grants the deployment process broad control over the host operating system's configuration and security posture.
[PROMPT_INJECTION]: The skill documents a server architecture that processes untrusted data in the form of JWT tokens, representing an indirect injection surface.
Ingestion points: The Resonate server ingests JWT tokens via the Authorization header as described in the configuration and troubleshooting sections of SKILL.md.
Boundary markers: The skill outlines the use of prefix claims to scope access, acting as a logical boundary, though it lacks explicit instructions for sanitizing or validating claim values against injection attacks.
Capability inventory: The server binary runs as the root user and has capabilities to perform network operations and write to local SQLite or PostgreSQL databases.
Sanitization: The skill relies on standard RSA signature verification to ensure token integrity, but does not detail further validation or escaping of the token's payload content before processing.

resonate-server-deployment