Competition Template Render Path is a downstream skill of ctf-sandbox-orchestrator for challenges where the decisive behavior lives in server-side rendering, template data or hydration handoff. It guides an agent through the render chain from route resolution to client takeover.
It is meant for CTF and lab environments that the orchestrator has already set up. The skill compares template-level gating, loader-level gating and backend handler enforcement, and packages the smallest request-to-render sequence as evidence. Replies are in Simplified Chinese unless English is requested.
What You Get
- A step-by-step workflow from route match to loader, template context, emitted HTML, hydration payload and client takeover
- A checklist of what to record on the first pass: host, route params, preview toggles, hidden fields, meta tags and inline data
- A way to compare template gating with loader and handler enforcement
- An evidence layout that keeps route side, render side and effect side together, with one success and one failure pair
Common Use Cases
- Explaining why a preview or hidden route renders content that the visible interface does not offer in a CTF lab
- Finding hydration JSON or inline data blobs that expose a branch the template hides
- Checking whether feature flags or tenant switches are enforced by the loader, the template or the backend handler
- Writing up a render-layer finding with a minimal request-to-render sequence
What Makes It Unique
Most web skills treat the page as an HTTP response. This one treats the render pipeline as the object of study: route selection, context construction and hydration are separate stages, and a finding has to say which stage holds the weakness. It also hands off to the bundle and source map skill when the work turns into recovering emitted bundles, so the scope stays narrow.
Under The Hood
The skill is a markdown workflow with one reference checklist and an agent metadata file that disables implicit invocation, so it only runs after the orchestrator routes to it. It ships no scripts. The instructions tell the agent to inspect requests and responses of the sandbox target, so the agent will issue web requests and run whatever commands the session allows. It writes nothing on its own beyond the evidence the user asks for.