Overwatch

{"notes":{"caps":"The facts serious enough to put the better bands out of reach, tested against the evidence. A cap only ever lowers the band, never raises it, and always names the findings that fired it.","against":"The strongest points against the band, each of them weighed before it was awarded.","observations":"An absence verified, a risk ruled out, a scope confirmed: what this audit established about the repository that is not a numbered finding, and that a reader would otherwise have to infer from the silence.","findings":"One row per finding: what is wrong, what proves it, and what the work costs — on one line. `verified` means the evidence carries a command and that command’s output; anything else lowers coverage and never flatters the band.","checklist":"One row per check: what was looked for, what answers it, and — where the answer is no — what the work costs, on one line. Every point in the total above is counted from these rows, so each one can be taken apart on its own.","not-verified":"The claims resting on something other than a command and its output, each with what would settle it. They are stated here rather than counted as passes.","convention-debt":"Rules this codebase states and nothing enforces: written down, agreed, and left to memory. Each row says where the rule is written and what is missing for it to hold on its own.","disagreements":"Where the figures the report states about itself differ from what its own rows add up to. The rows are what the numbers above were counted from.","fit":"Where the questions this audit asks did not suit this codebase: something that could not honestly be answered yes or no, or a real weakness with no row to sit in.","scope":"How much of the repository each territory read: the files in its scope, and the ones it did not get to. The band rests on what was read.","actions":"The work this audit surfaced, in the order that removes the most risk for the least effort rather than by severity label."},"floor":"floor this audit puts under it"}