NEW CAPABILITY 02 / 06
Eval scorecards for skills
Every collateral skill on this site is scored by a harness, and the results are published.
You can use this right now. It runs in your browser, on this site’s own data.
A skill that produces documents can be tested like code. This harness runs a fixed set of checks against the twelve collateral systems already on this site, and the scorecard below is its real output, failures included.
TRY IT / RUNS IN YOUR BROWSER
Use it here.
Loading the scorecard.
WHAT IT PRODUCES
The artifacts.
THE MECHANISM / SHOW THE SHAPE, NOT THE SOURCE
How it works.
- 01Define
Each check gets an id, a plain description of what it proves, and a severity.
- 02Run
The harness reads the generated files and their renders. It never edits them.
- 03Score
Each skill is scored on the blocking checks it passes.
- 04Publish
The scorecard is written as data and rendered on this page.
- 05Gate
A failing blocking check exits non-zero, so a build can refuse to ship.
WHAT IT READS
- The twelve skill definitions
- Every generated document and its rendered PDF
- The data files a document claims to be built from
THE RULES IT HOLDS ITSELF TO
- A check that cannot be made reliable is marked advisory, not passed
- Failures are published rather than fixed by loosening the check
- Page counts are read from the rendered PDF, not from the source
- Prices on a sheet are recomputed from the data file behind it
This is a new capability, built for this portfolio. It is not part of a client engagement, and nothing here is connected to a company’s internal system. The sample company, Apex Instruments, does not exist.