The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of high-privilege shell commands to manage the KASM environment.
Uses sudo docker exec kasm_db psql across multiple files (SKILL.md, references/bulk-operations.md, references/workspace-management.md) to directly modify the system database.
Employs ssh root@$SERVER in references/capacity-planning.md for system-level monitoring (e.g., free -h, docker stats).
Uses rm -rf and chown in references/backup-operations.md for filesystem management.
[PRIVILEGE_ESCALATION]: The skill provides templates and instructions to bypass security controls within workspaces.
Instructions in references/bulk-operations.md and references/config-templates.md show how to inject NOPASSWD: ALL into the /etc/sudoers.d/kasm-user file inside containers.
Enables privileged: true mode for Docker-in-Docker workloads in references/config-templates.md.
[EXTERNAL_DOWNLOADS]: The skill references the use of external tools for monitoring and data synchronization.
Recommends the use of sentinelctl for fleet monitoring in SKILL.md and references/capacity-planning.md.
Uses rclone for synchronizing profile data to external providers like Backblaze B2 in references/backup-operations.md.
[DYNAMIC_EXECUTION]: The skill facilitates the creation and execution of dynamic scripts.
Uses the docker_exec_config and first_launch fields to store and execute shell command chains (e.g., bash -c "...") when containers are started, as seen in references/config-templates.md and references/bulk-operations.md.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection due to its operational role.
Ingestion points: Ingests data from KASM API endpoints (get_kasms, get_images) and PostgreSQL database queries.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands in the retrieved data are present.
Capability inventory: Powerful capabilities including sudo docker exec, ssh, and curl are available across all management scripts.
Sanitization: There is no evidence of input validation or sanitization for data retrieved from external API or DB sources before it is processed or used in further commands.

kasm-admin