CyberRouter · Practical guides
AI security research & LLM workflow guides
Choose a task, follow a worked example, and check the result yourself. These guides connect security research and report quality with the practical details of running your own models.
Security research and reporting
Work from evidence, check AI-assisted drafts, and reproduce a small authorization example.
AI-Assisted Bug Bounty Reports: Check Every Claim
Review an AI-assisted report against the original observations, separate assumptions from evidence, and prepare a reproducible draft.
Read guide →AI Security Code Review: A Synthetic Authorization Example
Use a small authorization example to judge an AI review against a reference rubric, then inspect the references and cost of your own run.
Read guide →Context windows and MCP tools
Understand what uses the context window and check changes with concrete examples.
Fix Context Too Large Errors in LLM and OpenCode Requests
Find which part of a request exceeds the endpoint’s limits, then make a deliberate change to history, tools, retrieval, or output.
Read guide →RAG Context Budget: Calculate How Many Chunks Fit
Work out how much room is left for retrieved passages, including the overhead around each chunk.
Read guide →MCP Tool Token Budget: Count Schemas and Results
Account for the tool definitions and results your host actually sends to the model.
Read guide →Reduce MCP Schema Tokens: A Reproducible Example
Compare two downloadable tool definitions, reproduce their token counts, and check what a smaller schema does and does not establish.
Read guide →Provider keys and costs
Understand provider charges and gateway fees before connecting your workflow.
Start with a free tool
Build a Markdown report, calculate a context budget, or inspect MCP tool schemas in your browser. No account or provider key required.
Inspect the evidence behind CyberRouter
Read the CyberGym subset methodology, selected CVE capability scores, and a researcher’s account of a rewarded finding.
Research results and sources →