--seed reproduces a run again: A tool whose input has two or more properties now gets the same generated arguments on every run with the same seed, so a result can be replayed exactly. No check was added or removed: 120 checks, as in 0.0.7.
Simpler code behind every verdict: Every function is now under standard complexity limits, and lint keeps it there. Against 0.0.7, scout produces the same findings, scores and reports for the same server.
CodeQL on every pull request: scout, scout-reporting, scout-mcp, scout-action and the website are now analysed by CodeQL, weekly and on every change, alongside the existing dependency scanners.
scoutmcp.io has its own repository: The website is built from scout.github.io against scout's latest release, so its numbers always come from a real run.