The Agent Skills Directory

[COMMAND_EXECUTION]: The helper scripts utilize node:child_process to run the rig CLI for system inspection and sqlite3 for database management. These operations are limited to standard diagnostic and backup functions, with path validation and safe argument handling employed to prevent shell-related injection.
[INDIRECT_PROMPT_INJECTION]: The migrate-telemetry-state-0.5.9.mjs script ingests local telemetry and configuration data to manage state transitions.
Ingestion points: Telemetry sidecar files (.json) and instance configuration (config.json) in the OpenRig home directory.
Capability inventory: File writes (atomicWrite), directory management, and rig command execution.
Boundary markers and sanitization: The script performs strict JSON schema validation, timestamp checks, and identity verification before processing the data, mitigating the risk of instructions embedded in telemetry data influencing agent behavior.
[SAFE]: The skill instructions and scripts consistently promote safe practices, such as verifying state after each mutation, using integrity checks for backups, and avoiding the automatic deletion of local modifications. No network operations or unauthorized data exfiltration patterns were detected.

openrig-upgrade