This skill is a specialization of the ctf-sandbox-orchestrator skill for challenges about what a live container or pod is doing now, as opposed to what its checked-in manifest says. The agent separates manifest, image, startup, live mount, live route and live process, and maps the chain from host and proxy to container, mounted volume and consuming process.
The skill states that it runs only after the orchestrator has set sandbox assumptions.
What You Get
- A three-step workflow: map the live runtime, trace route and mount boundaries, report the runtime deviation
- A reference checklist for comparing intent with reality and for tracing who writes, mounts and reads each file
- A list of high-value deviations such as rendered secrets on shared volumes and init or sidecar output
- Hand-off rules to four sibling skills for control plane, routing, container escape and Linux credential questions
Common Use Cases
- Finding which process consumes a mounted secret or rendered config in a lab container
- Explaining why a route works only after startup because of an init container or sidecar
- Documenting how a virtual host and proxy resolve to a specific live container
What Makes It Unique
It starts from the position that checked-in manifests are intent and the live container is the evidence. Its checklist asks for the consuming process behind every mounted secret, and it flags sidecar and init output that is missed when only the main service is inspected.
Under The Hood
The skill is instruction text only: a SKILL.md, one reference checklist and an agent configuration file. It ships no scripts. The method involves running container and cluster inspection commands and reading mounted files, so the agent’s own shell tools do that work in the sandbox the orchestrator set up. The skill does not fetch pages or call APIs on its own.