The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to manage local infrastructure, such as just run-multitenant-local for building/deploying and kubectl for port-forwarding and process management.- [DATA_EXFILTRATION]: Extracts an administrative API token by reading logs from a running container (kubectl logs). While intended for user setup, this involves the agent accessing and displaying sensitive live credentials.- [COMMAND_EXECUTION]: Uses kubectl exec to retrieve a server certificate from within a container and writes it to a temporary file on the host (/tmp/duckgres-server.crt).- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests and processes unvalidated data from service logs. A compromised service could output logs containing instructions designed to influence the agent's behavior.
Ingestion points: Output of kubectl logs in SKILL.md
Boundary markers: Absent
Capability inventory: kubectl, just, pkill, and filesystem write operations
Sanitization: None present

multitenant-up