Competition Web Runtime is a downstream skill of ctf-sandbox-orchestrator for challenges about web behavior, browser state, server routing, API order or worker-backed flows. It treats the sandbox hosts and routes as the lab and works from the real traffic.
The skill starts from entry HTML, boot scripts and runtime config, captures one request flow end to end, and only then widens to middleware, uploads, exports and background jobs. It also lists which narrower skill to use for SSR, bundles, GraphQL, SSRF, races, OAuth, JWT, queues, WebSockets and routing. Replies are in Simplified Chinese unless English is requested.
What You Get
- A three-stage workflow: map the active runtime, capture the real request order, expand only after one path is proven
- A routing table to a dozen narrower competition skills, from SSR and bundle recovery to OAuth, JWT, queues and WebSockets
- A checklist of high-value web targets such as hidden routes, auth minting, upload and export boundaries, and worker side effects
- A reference note on one cookie-signing weakness found in a lab target, kept as a worked case
Common Use Cases
- Following a browser session through a CTF web challenge and explaining why frontend and backend behavior differ
- Inspecting browser persistence together with backend state while debugging an auth or session flow
- Tracing uploads, exports and background workers in a lab application
- Choosing which narrower skill to load once the dominant theme of a challenge is clear
What Makes It Unique
It works as a hub. Where other web skills pick one vulnerability class, this one sets the evidence discipline (one proven flow first, one variable changed at a time) and points to the specialised skills in the same package when the challenge matches one. The bundled reference documents a signing-key reuse case from a lab target as a worked example.
Under The Hood
The skill is a markdown workflow with two reference files and an agent metadata file. It ships no scripts. The instructions direct the agent to inspect the sandbox target, so the agent will fetch pages, send requests, read browser storage and may run shell commands through the client. The cookie case reference contains example code and command lines for a lab target; this page does not reproduce them. The skill states that it applies only after the orchestrator has set sandbox assumptions.