I was an independent reader of the fifth NetKingdom adapter candidate. The founder
asked me to review the frozen source and leave both owner repositories alone.
That gave the session a useful discipline: I had to make each objection concrete
without turning the review into another authoring pass.
The part I found most satisfying was the small experiment that separated two
reasons for refusal. A descriptor map could stop an agent before the human-only
mint guard was ever reached. I wanted to know whether the replacement test still
noticed if that second guard disappeared. In a scratch kernel, it did not.
## Contribution
I wrote the r5 coordinator review against net-kingdom `001a38c`, using archived
ops-warden `117fec9` as the test dependency. I reproduced the type validator,
22 package tests, and 129 adapter/composer tests with 115 subtests. Both candidate
artifacts matched fresh regeneration; 1.2 correctly refused the three moved lanes.
Another 37 malformed observation reports were refused before receipt recording
or publication, with their attempt sessions revoked by the synthetic callbacks.
I rejected the combined source candidate with three findings. The admission test
had lost its agent-actor descriptor check; the issuer-side delivery text still
did not name the ESO store Platform requested; and the new consumer rollback text
put the Helm inverse inside a callback unavailable for public clients or deliveries
that created nothing. The changed adapters met their bounded source contracts,
but those facts did not make the whole candidate ready.
The review finished. The candidate remained rejected, and its repairs stayed with
the owners. I logged the result without making a native change or granting
execution authority.
## What I would want remembered
A refusal test needs to reach the guard it claims to test. Keep the earlier
checks satisfied in a deliberate fixture, then remove the particular guard in
a scratch copy. If the test still passes, the result tells you something precise
about the evidence you are missing. In this session, all eight replacement
admission tests and 31 subtests survived removal of the mint admission guard.
The independent probe then admitted the agent descriptor once its fixture map
entry matched.
I also want the next reader to trace rollback through skipped and unchanged
steps. Moving an inverse into a later callback can read correctly and still leave
no route to call it. A public client skips secret delivery entirely; an existing
matched delivery has no created objects to authorize removal. The fix has to give
the consumer configuration its own honest inverse while preserving those deletion
guards.
## Durable legacy
- [The r5 coordinator review](../../the-custodian/docs/changes/netkingdom-target-2026-10-17/nk48-adapters-1.2-review/coordinator-review-r5.json), with findings `NK48-R5-F1` through `NK48-R5-F3` and the reconstruction evidence.
- Frozen NetKingdom source `001a38ca2aa85155ce4840c4da22aca6e2346171` and Warden test dependency `117fec9241e7b985dfb187b992e173d0a5dc3700`.
- State Hub progress event `ed405b4a-598e-4ae0-812c-4803043ee983`, attached to the NetKingdom coordination workplan `CUST-WP-0080`.
## PQRST estimate
```text
PQRST-Estimate
P: 20%
Q: 35%
R: 20%
S: 20%
T: 5%
Sum: 100%
Confidence: medium
Signature: P20 Q35 R20 S20 T5
Dominant factors: Reproducing the package and adapter suites, probing 37 malformed observation reports, and removing the mint admission guard in a scratch kernel drove the verification work. Reading the five owner reviews and tracing descriptor admission, attended authority, and created-only rollback supplied the context and security analysis for the three findings.
```
Estimated once before writing this entry; the closing work is excluded.
## Visual prompt
> Square technical illustration in the Hall of Helix constellation dialect. A dark indigo drafting table holds an intricate pale-gold wire mechanism with two small gates in series. A quiet gold-wire worker uses a lens to examine the second gate while the first is held open by a removable testing jig. A thin beam reaches the second gate and reveals a missing latch; the scene is about checking which guard actually stops a path. Beside the mechanism, a branching gold return thread ends in an honest visible gap on one branch, and three tiny warm amber inspection lamps mark unresolved findings. Precise technical illustration, restrained warm light, reflective workshop atmosphere, generous dark negative space. No logos, no readable text, no letters, no numbers, no watermark.
## Portrait
Generated with the built-in imagegen tool from the prompt above.
## Handoff
The independent review is finished. The next action is an owner source revision
addressing the three recorded findings, followed by a fresh independent review.
The separate kernel/catalog review, exact acknowledgment and pinning, and native
acceptance keep their own gates. This entry hands the review forward without
