Capability, qualification, and machine-checked claims
August 26, 2026 ยท View on GitHub
Two admissions answer two different questions
A release admission and a workflow admission answer different questions. The release admission covers the exact product release. The workflow admission covers one exact sealed bundle, application, environment, runtime, input, identity, effect, and policy contract.
Flow enters Production only through an active signed, expiring, and revocable release admission. If the admission is missing, expired, revoked, or bound to a different release, Flow is not actively admitted. Read the current admission-derived status and the capability and qualification matrix.
What is qualified today
The browser path runs record, compile, policy-check, deterministic
replay, refusal, and report generation in CI. Windows UIA, native macOS, native
Linux, and RDP each have retained 3/3 accepted task evidence with independent
effects or oracles. Citrix has a dedicated exact-window backend and a retained
3+3 code-readiness record; an exact ICA/HDX deployment receives its own counted
qualification record rather than inheriting RDP or stand-in evidence. Each new
third-party application is similarly qualified against its controls and effect
oracle. The workflow-program IR adds parameters, branches, loops, effect
verification, and governed recovery on the same runtime. DESIGN.md has the
module contracts;
design/WORKFLOW_PROGRAM_IR.md
describes the program IR, and L1_INTEGRATION.md
covers feeding layered clinical-data platforms.
The integrated status of the engine, browser, desktop, remote-display, safety,
GUI, hosted, and deployment surfaces is published in
PRODUCT_STATUS.md. Security reviewers should
start with ENTERPRISE_ARCHITECTURE.md,
which maps screenshot/credential flows, cryptographic guarantees, hosted
boundaries, and unmet controls.
Machine-checked claims
Product claims are enforced by CI. Every registered claim is tiered and mapped
to the specific tests and benchmark artifacts that back it in
../claims.yaml. CI runs scripts/validate_claims.py, which
fails the build whenever a claim's tier outranks its strongest evidence and
regenerates VERIFICATION.md, the claim-by-claim
verification report, from the registry, so the adjectives in the README cannot
quietly rot.