How to use the Competition WebSocket Runtime agent skill

What the skill makes the agent do

The skill tells the agent how to analyze realtime connections. It maps the handshake (path, headers, cookies, query parameters, auth token and upgrade response), separates connection setup, subscription messages, keepalives, server pushes and reconnect logic, and records the message schema and channel identity for each frame.

The goal is to link each frame to the state change it causes, such as a rendered UI update or a stored value, and to reduce the result to the smallest handshake-plus-frame sequence that reaches the decisive change.

• Maps the WebSocket or SSE handshake, including auth material sent during setup
• Decodes frames and separates subscribe, ack, heartbeat, push and terminal messages
• Traces subscription state and reconnect behavior
• Explains how live frames change rendered or persisted state

Install the skill: download the ZIP and unzip it

Download the skill ZIP from the directory page. Unzip it so the folder named competition-websocket-runtime, containing SKILL.md, agents/openai.yaml and references/websocket-runtime.md, sits in one of these locations:

For a personal install, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.

In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/competition-websocket-runtime/
• Project: .claude/skills/competition-websocket-runtime/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Install the parent skill first

This skill is a downstream helper. Its source states that it requires the parent skill ctf-sandbox-orchestrator to be installed and active, and that the parent must already have established sandbox assumptions before this skill is used. Without the parent, the skill tells the agent to return to ctf-sandbox-orchestrator first.

Trigger the skill with natural requests

Agents load a skill when your request matches its description. This skill's description lists the kinds of requests it covers. Ask in those terms after the sandbox context is set up through the parent skill.

Examples of requests that match its description:

• Inspect the WebSocket handshake for this lab and list the headers and cookies it sends.
• Decode these frames and tell me which message subscribes to which channel.
• Trace the subscription state and explain what happens when the connection drops and reconnects.
• Show how the live frames change what is rendered on the page or saved to storage.
• Explain the auth material sent during realtime setup in this SSE stream.

What to expect from the output

The skill asks the agent to keep evidence grouped in one chain: setup details (request shape, tokens, cookies, channel identity), frame details (type, topic, payload schema, order, reconnect behavior) and effects (UI update, storage change, server action). Responses are to be given in Simplified Chinese unless you explicitly ask for English.

The skill also warns against treating keepalive traffic as business logic and against listing frames without showing which one changes state.

• Handshake and auth details, grouped together
• Frame order and a minimal replayable sequence
• The specific frame that changes state, and the effect it has

Safety notes: scripts and network access

The skill's files list no scripts to run and no network use. It reads its reference file, references/websocket-runtime.md, and names no external software, libraries or API keys.

The skill is designed to inspect auth material and reproduce handshake-plus-frame sequences. Use it only against systems you own or have written permission to assess, such as a CTF or security training sandbox. It is not meant for one-off HTTP requests, which its description places outside its scope.

• Runs scripts: no
• Uses network: no
• Only use it in authorized sandboxes or on your own application

Tips for better results

Set up the sandbox assumptions with ctf-sandbox-orchestrator before asking for frame analysis. Give the agent the captured handshake and frames if you have them, since the skill works from the traffic you provide.

If the problem is only protocol decoding without any UI or app-state link, the skill itself suggests switching to a tighter protocol skill.

• Run the parent skill first in the same session
• Provide the handshake and a few representative frames, including the reconnect
• Ask for the smallest sequence that reproduces the state change