The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to access and use a sensitive private SSH key file located at ~/.ssh/cloud-weaver. This key is used for authentication when connecting to the remote virtual machine.\n- [PRIVILEGE_ESCALATION]: The deployment script scripts/deploy-hermes.py performs operations on the target host using sudo, including creating system directories, modifying file ownership for database data, and running Docker Compose.\n- [COMMAND_EXECUTION]: The skill executes local system commands via a Python script that uses subprocess.run to invoke ssh, scp, and docker compose. It also expects the agent to execute the Python script itself with various command-line arguments.\n- [EXTERNAL_DOWNLOADS]: The deployment stack downloads the postgres:17-bookworm and devlikeapro/waha:latest Docker images from Docker Hub. The use of the latest tag for the WAHA image introduces a dependency on an unversioned external resource.\n- [INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided input for parameters like env_name and api_port. Although these are subject to validation checks, they represent a vector for potentially malicious data to reach the command line.\n
Ingestion points: User-supplied values for env_name, public_ip, and api_port collected in SKILL.md.\n
Boundary markers: The instructions define validation rules (e.g., [a-z0-9_]) which are implemented as regex checks in the deployment script.\n
Capability inventory: The skill has the ability to execute shell commands locally and remotely (via SSH), write to the local file system, and transfer files to remote hosts.\n
Sanitization: Input validation is performed using Python's ipaddress module for IPs, regex for naming, and integer range checks for port numbers before the values are used in deployment logic.

cloud-weaver-hermes