Expand description
The five-dimension benchmark scorecard — truth, feedback, change, uncertainty, workflow — plus the data-quality and token-economy blocks that qualify them.
This is the only implementation of the arithmetic. It previously
existed three times over: asd scorecard, the scorecard MCP tool, and
the /api/v1/scorecard handler each carried their own copy of the same
loop and the same formulas. That is the shape of defect this codebase has
already paid for once — see the calibration-table axis inversion in
CLAUDE.md, where every unit test passed because each test encoded the same
wrong assumption as the code it covered. Three copies of a formula are
three places for it to drift, each with its own tests agreeing with it.
What stays with the callers is phrasing and envelope, not computation: the CLI renders a table and keeps a trend snapshot, the HTTP handler wraps the payload in a timestamp, and each composes its own advice string for the “nothing matched” case, because “try broadening –scope/–paths” is good counsel in a terminal and meaningless over HTTP. Everything that produces a number lives here.
Structs§
Constants§
- TOKEN_
ECONOMY_ NOTE - The disclaimer that must travel with
TokenEconomywherever it is rendered. Kept as a constant so no caller can quietly ship the ratio without the caveat that it is an internal estimate.
Functions§
- compute
- Score the indexed symbol set.
- score
- The formulas, isolated from the gathering so they can be tested against hand-built metrics without an engine.
- sparse_
note - The sparse-ledger caveat. Identical wording across every surface, so it lives with the threshold that triggers it rather than being retyped alongside each renderer.