Console and Wire State Queries Read the Live State Graph, Not a Synthesized View

Claim

On the running eOS runtime platform (DGD 1.7.9 plus the eOS-kernellib kernel layer), state queries at both operator surfaces answer from the live state graph: operator console verbs enumerate current subsystem state over telnet, and authenticated HTTP reads return current record state over the wire. A mutation committed by one operation is visible to the very next query -- a console bind shows up in the next identity show, a registrar approval shows up in the next query-approved-registrars, an agent suspension reaches an open server-sent-event stream within one poll interval and refuses the agent's next authentication ceremony. No restart, no redeploy, no debugger attachment, and no separately-maintained introspection API sit between the query and the state it reports.

Grounds

This is an Empirical Observation. The grounds are three machine-asserted runs captured 2026-07-18 against the same platform build.

Setup (verifiable from repo state):

Run 1 -- the ten-verbset console drive (module-less):

The console-drive smoke script boots the platform headless with vault-app and upgrade-cascade deployed, logs into the real telnet console, and drives ten verbsets whose every expectation is a regex assertion against live verb output. All ten passed. The introspection reads exercised, each read returning current state and each read-after-mutation reflecting the mutation:

Run 2 -- the identity-recovery console verbset (crypto boot):

The console-drive smoke script run against the identity-recovery verbset, 10/10. The introspection-relevant sequence: identity mint 2 creates a record; identity bind <uuid> <challenge> <cdj> <ao> verifies a foreign-generated attestation and binds a passkey; the very next identity show <uuid> returns identity: credentials: 3 -- the two recovery-code rows plus the passkey the previous verb just bound. The read reports the record as it is after the atomic bind, with no intermediate refresh step.

Run 3 -- the composite example's wire surface (crypto boot):

The composite-app example's smoke run, 38/38 sentinels over real TCP. The introspection-relevant sentinels:

What the evidence establishes:

What the evidence does not establish:

The runs are sequential probes; no measurement here adjudicates read latency under load, concurrent-query behavior, or the cost of the reads. The wire read covers the own-agents scope only -- it does not demonstrate an unbounded cross-principal query surface (and the capability posture forbids one). The verbset assertions are regex expectations against verb output, not full-output captures; only the most recent run's raw transcripts survive locally.

What Would Revise It

Sources

Relations