The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses sensitive private SSH keys.
It identifies and utilizes SSH private keys located at ~/.ssh/cw-${REPO_NAME} for remote authentication.
[COMMAND_EXECUTION]: The skill executes various shell commands and scripts locally and on remote targets.
It runs local Python and Bash scripts included in the skill package.
It initiates ssh connections to execute Docker and system status commands on remote virtual machines based on user-provided IP addresses.
[PRIVILEGE_ESCALATION]: The skill utilizes elevated permissions on target systems.
The diagnostic scripts and manual commands provided in the instructions use sudo on the remote host to access Docker information and system logs.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to malicious instructions embedded in the data it processes.
Ingestion points: Reads and processes output from remote ssh commands and HTTP responses from health check endpoints.
Boundary markers: None; there are no instructions for the agent to ignore or delimit embedded commands in the output.
Capability inventory: The skill can write to the local file system (via --output flags) and execute shell commands.
Sanitization: None; the agent is instructed to interpret the output from these external probes directly.

cloud-weaver-monitor