I2652 — The UI shows what the knowledge base holds: the candidates awaiting review and whether the deriver still writes, read and never acted on

Give the queue the work leaves behind a screen: what awaits a reviewer, what it is about, and whether a derivation still follows every closed episode, so that the result of a review is seen where every other state of the delivery is.

DONE wave 2 · p1 · implementation profile · parallel safe

Part of development-runtime — The runtime has a development surface.

Objective

Why

The owner, 2026-10-13: bring /evolve into Majordomus as a derived skill that improves the know-how in .ai/, with its result visible on /ui and on Pages; no new systems, only connect, derive, project. The skill is another session's; this is the half a reader sees. Every organ exists: knowledge_base.status and knowledge_base.candidates answer the questions, and no screen asks them.

Current state

Measured on master 3442ce68c5 on 2026-10-13: the UI has thirteen screens and none names the knowledge base; src/ui mentions it only in passing in roadmap.rs and deployments.rs. The candidates are read from the command line, MCP and HTTP alone.

Desired state

A screen at /ui/knowledge reads knowledge_base.status and knowledge_base.candidates as readings and draws: how many records await review, on this branch and unattributed; the deriver's judgement in words with every finding; the subjects the records are about; and the newest candidates, bounded, each leading to its record, with what is left out counted. The review is named as the commands a person runs, as text. The screen stands in the navigation, in the published snapshot and under the UI's latency budget.

Scope

apps/majordomus-cli/src/ui/knowledge.rs
apps/majordomus-cli/src/ui/mod.rs
docs/MAJORDOMUS_UI.md
test/cases/3003_the_knowledge_screen_answers.sh
.ai/repo/project/issues/I2652.yaml

Out of scope

Promoting or rejecting a record from the screen: a screen reads and never executes (ADR 0142), and promotion is a person's act
The evolve skill itself and its projection as a command, which are another session's
A panel on the dashboard and a link from the Command screen, which follow once the application shell another session is building has landed
A change to what knowledge_base.candidates answers

Dependencies

I2486DONE No screen waits for a source: a page shows the latest reading with its age, read again behind the reader

Acceptance criteria

/ui/knowledge answers from knowledge_base.status and knowledge_base.candidates and names both as its sources with their age
A queue longer than the screen's bound is drawn newest first to the bound and says how many records it left out
An empty queue and a checkout where the deriver never ran are said in words, never as an empty table or an empty cell
The page carries no form and no button: the review is the commands a person runs, shown as text
No handler under src/ui executes a capability on the request, which case 1096 holds, and the route is within the UI's latency budget, which bench --check holds
Every screen's navigation leads to it

Validation

bin/majordomus plan validate
cargo test --manifest-path apps/majordomus-cli/Cargo.toml --lib ui::knowledge
MJ_TEST_JOBS=2 bash test/run.sh 3003_the_knowledge_screen_answers 1096_no_screen_waits_for_a_source
majordomus bench system.http. --transport system --profile ci --no-write --check

Evidence required

knowledge_screen_unit_tests
knowledge_screen_case_3003
knowledge_screen_within_budget

Evidence

covers	type	command	result	at commit
knowledge_screen_unit_tests	test	cargo test --manifest-path apps/majordomus-cli/Cargo.toml --lib ui::knowledge	test result: ok. 4 passed; 0 failed at 09c334079f5	09c3340
knowledge_screen_case_3003	test	MJ_TEST_JOBS=4 bash test/run.sh 3003_the_knowledge_screen_answers 1096_no_screen_waits_for_a_source	ok 3003_the_knowledge_screen_answers 2s; ok 1096_no_screen_waits_for_a_source 2s (0 skipped) at 09c334079f5 with the executable of #1304; delivered by #1090 (f0696ec535a)	09c3340
knowledge_screen_within_budget	test	majordomus-cli bench system.http. --transport system --profile ci --no-write --check	budget: system.http.ui.knowledge warm p95 399.6us of 100000us PASS, system.http.live.ui.knowledge warm p95 717.1us PASS, 'bench --check: within budget'; the run exits 10 for a baseline regression of system.http.openapi (p50 +582%), not this route, which is NEW to the baseline; at 09c334079f5	09c3340

Timeline

started: 2026-10-16T16:43:42Z
verified: 2026-10-16T16:43:47Z
completed: 2026-10-16T16:43:58Z

Those three fields, the evidence above and the state of the dependencies are all the status is made of. There is no status field to disagree with them.

Canonical record: .ai/repo/project/issues/I2652.yaml. Read it back with majordomus plan show I2652.