Matrix · 04 of 08 · Scorecard

The Airframe scorecard: ten dimensions, measured today

Re-runnable: hangar/tools/airframe-scorecard/scorecard.py. It measures readiness against the full contract, so most of the gap is artifacts that do not exist yet, not things that are broken. The amber tick on each bar is the A+ line at 97.

The Airframe scorecard: ten dimensions, measured today Matrix of ten Airframe AI-friendliness dimensions with the measured baseline score, what the measurement found, and what an A+ requires, with an overall baseline of twenty-seven out of one hundred and one of fourteen acceptance checks passing. DIMENSIONSCOREWHAT THE MEASUREMENT FOUNDWHAT A+ CHANGES1 Discoverability35no AGENTS.md or contract bundle; 6 of 13 XRDs catalogedAGENTS.md, a generated contract bundle, a capabilities tool2 Schema precision4512/234 nodes described; 0/68 objects strict; typos passschema-first, strict, 95% described, discriminated components3 Component contracts422 of 25 env entries hard-code a derived namedeclared outputs, fromComponent, verify checks4 Pre-merge validation22no validate; ground envs commit straight to mainairframe validate, six layers, required checks, dead-end rules5 Write safety134 of 4 live files mix human and machine-owned keysone file one owner, a base layer, field risk classes6 Observe and verify21Ready and Synced only; no reason codes or verifya closed reason list, verify contracts, describe7 Docs for agents39true quickstarts; none validated or executableexecutable walkthroughs, YAML validated in CI8 Interaction surface27Tower UI only; no plan, apply or verify APIAppSpec, planner, airframe tools, resumable9 Safety integration20tiers reach paths, not fields; extraManifests is openfield-level scope; escape hatches denied by default10 Hygiene46lint passes; no chart tests, no chart CIchart tests, CI, a scorecard gate, a compatibility matrixOverall 27 / 100 · 1 of 14 acceptance checks · A+ needs 97 and 14 of 14LEGENDOverallTarget stateMeasured today

Principle honored

Measured, not felt. Every check is automated, so "A+" is a result the scorecard prints, not an opinion.

What changes

  • The biggest gaps are also the cheapest: AGENTS.md, a contract bundle, validate, outputs.
  • Two real bugs found by running the chart: typos pass silently, and configuring before an image renders image ":".
  • Baseline committed: tools/airframe-scorecard/baseline-2026-09-26.json.

Limits

The scorecard cannot judge whether an agent would actually succeed; the parachute sentence (one request asking an agent to set up a new Python app named parachute, end to end), run as a Preflight case, does. Use both.

Hangar · Autopilot · plan