Web UI Roadmap pages

  - attempt 1 builder, session 56246e56-e26e-4339-b25f-01a1ae5c4f31: pass (said done)
  - attempt 1 verifier, session d404ffa3-3403-449c-87fe-01a1ae5ff5d3: pass (Evidence written by the verifier)
- web-change-set-page (step build/build): passed: attempt 1: verifier passed
  - attempt 1 builder, session fb8f6edf-65a6-42d4-8cad-01a1ae603308: pass (said done)
  - attempt 1 verifier, session b57f3891-c212-43b2-9ca3-01a1ae63b03c: pass (Evidence written by the verifier)
- web-roadmap-page (step review/review): open
  - attempt 2 fixer in cycle 2, session b85c4d10-19a0-47f0-b9de-01a1ae7bda4f: pass (said done)
  - attempt 2 verifier in cycle 2, session 65b4429b-e60e-4383-91f6-01a1ae8273de: pass (Evidence written by the verifier)
- web-change-set-page (step review/review): open
  - attempt 2 fixer in cycle 2, session 42495fce-1f7f-4e07-9e27-01a1ae82bf7c: pass (said done)
  - attempt 2 verifier in cycle 2, session 31bbabd7-386c-476a-97b7-01a1ae86e0fa: pass (Evidence written by the verifier)
  - attempt 5 fixer in cycle 3, session 0b0d7fe8-28ec-444e-8399-01a1ae92907b: pass (said done)
  - attempt 5 verifier in cycle 3, session 49cfd641-2f52-4285-b767-01a1ae9457dd: pass (Evidence written by the verifier)

## Checks

- compliance (step review/review): passed after 1 cycle
  - cycle 1: passed, check session 4bd6a5fd-37ec-4b6c-b08b-01a1ae6f3fd7 pass (said done)
  - round 1 (cycle 1): 2 findings, 0 confirmed
    - f1 minor at internal/web/roadmap_test.go:89 on web-roadmap-page: The page's order is checked against a hard-coded list, and the roadmap operation's response is only checked for 3 groups with kinds at [0] and [2]; the change sets' order within each group is never read from the operation, so 'the same groups and order as the roadmap operation' is asserted indirectly rather than derived from it.
    - f2 minor at internal/web/roadmap_test.go:101 on web-roadmap-page: Each change set's state is checked as 'rz-pill--<state>' appearing anywhere in the body, not on that change set's row, so a row with the wrong state pill would pass as long as another row has that state.
- quality (step review/review): passed after 3 cycles
  - cycle 1: failed, check session 579cde3b-c2bc-4899-be0d-01a1ae6f8223 pass (said done), critic session c4354d31-5bf4-4dbf-a6ca-01a1ae79a339 pass (said done)
  - cycle 2: failed, check session 00b4794a-7f2c-4265-81f2-01a1ae8a65a8 pass (said done), critic session e4d9a306-38c8-4de7-a63f-01a1ae91609c pass (said done)
  - cycle 3: passed, check session 6d941432-2e4b-476e-a8ef-01a1ae95defb pass (said done)
  - round 1 (cycle 1): 5 findings, 1 confirmed
    - f1 major at internal/web/change.go:203 on web-change-set-page: The anonymous run.ls struct declares StartedAt with no json tag, but run.ls emits "started_at". encoding/json's case-insensitive match does not ignore underscores, so StartedAt is always empty (confirmed with a decode of {"started_at":...}). The time.Parse on line 223 never succeeds, and every run row on the change set page has an empty Updated column. Add `json:"started_at"` (and tags on the other fields), and make TestChangeSetPage assert the run's time (2026-10-24 11:00). — confirmed: internal/web/change.go:203 decodes run.ls into an untagged StartedAt field, while api.RunInfo (internal/api/run.go:174) emits json:"started_at". encoding/json matches field names case-insensitively but does not strip underscores, so StartedAt stays empty, time.Parse fails, and every run row's time is blank.
    - f2 minor at internal/web/change_test.go:56 on web-change-set-page: The test asserts neither the run row's time nor the history entries' times, so the started_at decode bug above passes unnoticed.
    - f3 minor at internal/web/roadmap.go:156 on web-roadmap-page: The roadmap fragment's Data swallows a loadRoadmap error and returns an empty view, so a failed API read re-renders the live section as 'Nothing is planned' instead of an error, while the full page returns a 500 for the same failure.
    - f4 nit at internal/web/change.go:241: h.Type[:1] panics on an empty lifecycle type. It is unlikely, but a guard costs nothing.
    - f5 nit at internal/web/roadmap.go:29: Production code imports net/http/httptest to call the API in-process. It works, but a small in-package ResponseWriter, or calling the operations directly, would keep test helpers out of the binary.
  - round 2 (cycle 2): 3 findings, 1 confirmed
    - f1 major at internal/web/change.go:116 on web-change-set-page: showRef now always reads with view=what-if:<cs>, but the API refuses a what-if view of a change set that is not draft, proposed or accepted. For a landed change set it returns conflict: "change set 01M6K044... is \"landed\", not draft, proposed or accepted, so it has no what-if view" (checked with the built binary: `rhiza show <leaf> --view what-if:<landed cs>`). loadChange drops every showRef error, so a landed, withdrawn or rejected change set's page shows '0 leaves, 0 with evidence', no Serves and no questions. The page's own Depends on links lead to such pages: the test fixture's dependency is a landed one. Use what-if only for in-flight states (cs.State), fall back to as-is or to-be for landed ones, and test a landed change set's page. — confirmed: internal/web/change.go:116 showRef always reads view=what-if:<cs>, and internal/api/read.go:329 refuses a what-if view of a change set in any state but draft, proposed or accepted. With this branch's build, `rhiza show darin-haener --view what-if:01M5SWRGEM7Z4C61JMG9EZRQFX` (a landed change set) returns that conflict. loadChange skips every showRef error, so a landed, withdrawn or rejected change set's page shows no leaves, no evidence, no Serves and no questions. The Depends on links lead to exactly those pages, and change_test.go only tests an accepted change set's page.
    - f2 minor at internal/web/change.go:121 on web-change-set-page: Depends on now comes only from the roadmap operation, which lists only in-flight change sets. A landed change set's page (or a draft's, if roadmap omits drafts) therefore always shows 'Depends on —'. Each page view also folds the whole roadmap to find one change set's depends_on.
    - f3 nit at internal/web/templates/roadmap.html:7: The backlog aside still renders '{{len .Rows}} change sets', so a backlog of one reads '1 change sets'. The change page's '{{len .Data.Runs}} runs' has the same problem.
  - round 3 (cycle 3): 1 finding, 0 confirmed
    - f1 minor at internal/web/change.go:119 on web-change-set-page: A withdrawn, rejected or superseded change set now reads the as-is view, where the leaves and questions it created never landed, so its page shows '0 leaves, 0 with evidence' and no questions, the same as before the fix. Only the landed case is tested. That may be acceptable for sets that never landed, but the page says nothing about why the list is empty.
- qa (step review/review): passed after 3 cycles
  - cycle 1: failed, check session 3eb24189-72b9-46d3-b820-01a1ae6fc682 pass (said done), critic session 7452807b-bb05-438d-8154-01a1ae79df3a pass (said done)
  - cycle 2: failed, check session 0ebbd179-5ce0-4894-bf2d-01a1ae8ad2ff pass (said done), critic session 85706367-f4e2-40e6-98f2-01a1ae919d0c pass (said done)
  - cycle 3: passed, check session 4b0924b9-1453-4e93-932e-01a1ae964e12 pass (said done)
  - round 1 (cycle 1): 5 findings, 3 confirmed
    - f1 blocker at internal/web/change.go:118 on web-change-set-page: showRef calls /api/show with no view, so leaves, their Evidence and their questions resolve only when the daemon's repo happens to have a current change set. With none (the daemon on a clean main tree), the view is as-is. Every proposed or accepted change set's leaves then 404 and are silently dropped: on the built daemon, /changes/01M6PFH625P3BYQAB8P6J38BBT showed '0 leaves, 0 with evidence' while change.show lists 2. The test passes only because its fixture leaves the shown change set current. Read with view=what-if:<change set>, and test with no current change set. — confirmed: showRef (internal/web/change.go:118) calls /api/show with no view, so the view follows the effective change set (ShowRequest.View default). With no current change set the view is as-is, and an unlanded leaf is not there: `rhiza show 01M6PFKANQ23RTM5QSR6H1CGBX --view as-is` returns not_found, but `--view what-if:01M6PFH625P3BYQAB8P6J38BBT` finds it. loadChange skips leaves on error (change.go:169-171), so every unlanded change set's page shows no leaves, Evidence or questions. TestChangeSetPage passes only because newChange leaves cs current. The leaf's core content goes missing on the daemon's normal path, so this is a blocker.
    - f2 major at internal/web/change.go:145 on web-change-set-page: Depends on reads only the depends-on links added by the change set's own ops. Dependencies are normally added by another change set, such as the milestone plan, so every change set I opened showed 'Depends on —'. The roadmap row for the same change set says 'depends on 2 change sets', and /api/roadmap gives depends_on [01M6PFEX…, 01M6PFFF…]. Use the model's depends-on links (the roadmap operation's depends_on or changeset.DependsOn) instead. — confirmed: change.go:144-158 builds DependsOn only from cs.Changes.Links, which holds the links this change set's own ops add. `rhiza change show 01M6PFH625P3BYQAB8P6J38BBT --json` has no depends-on link in changes.links, yet `rhiza roadmap --all --json` gives it depends_on [01M6PFEX…, 01M6PFFF…]: the milestone plan added them. The leaf's statement asks for what the change set 'depends on', and the page shows none for it, so this is major.
    - f3 major at internal/web/roadmap.go:108 on web-roadmap-page: Every group that is not the backlog becomes a milestone. The roadmap operation returns milestone, milestone, backlog, umbrella('areas'), but the page shows 'areas' as a third milestone ('3 milestones · 16 change sets · 1 in backlog') and puts it before the backlog. So it does not show the groups in the order rhiza roadmap gives them, and it mislabels the umbrella group. — confirmed: In loadRoadmap (roadmap.go:108-146), every group whose kind is not 'backlog' goes into Milestones, with the title 'Milestone' or the intent's statement, and the template renders all Milestones before the Backlog. The roadmap operation returns milestone, milestone, backlog, umbrella(areas). The page therefore labels the umbrella as a milestone, counts it in 'N milestones', and moves it ahead of the backlog. That breaks the leaf's 'as rhiza roadmap groups them', so this is major.
    - f4 minor at internal/web/change.go:204 on web-change-set-page: The run row's untagged StartedAt field does not match run.ls's started_at key (encoding/json ignores the underscore), so a run's time is always blank: the ship-change run row showed no Updated time.
    - f5 nit at internal/web/templates/roadmap.html:14: Fixed plurals: the backlog aside renders '1 change sets', and change.html:21 renders '1 runs'.
  - round 2 (cycle 2): 2 findings, 1 confirmed
    - f1 major at internal/web/change.go:114 on web-change-set-page: showRef always reads with view=what-if:<change set>, but the what-if view refuses a change set that is landed (or withdrawn, rejected or superseded): /api/show answers 409 'is "landed" ... so it has no what-if view'. loadChange drops every node that fails to load, so every landed change set's page shows '0 leaves, 0 with evidence', Serves — and no questions. On the built daemon, /changes/01M6J965BRJ2ZTDT7Y311PM96G showed 0 leaves while change.show lists 3, and 01M6FENG43982SZMA5Q9VBGEVZ showed 0 while it has 1. Use what-if only for draft, proposed or accepted change sets, and the default (as-is) view otherwise, and add a landed change set's page to TestChangeSetPage, which today renders only an unlanded one. — confirmed: showRef (internal/web/change.go:114-117) always reads with view=what-if:<cs>, and the what-if view refuses a change set outside draft/proposed/accepted (internal/api/read.go:327-329). With the built binary, `rhiza show 01M6J9NTD1SF9Z8B72JJ0MV1HG --view what-if:01M6J965BRJ2ZTDT7Y311PM96G` returns conflict '... is "landed" ... so it has no what-if view'. loadChange skips every node whose showRef fails (Serves at change.go:149, leaves at 180-182), so a landed change set's page shows no leaves, Evidence, Serves or questions. Those pages are reached from the page's own Depends on links, so the change set page is wrong for a normal state: major.
    - f2 nit at internal/web/templates/roadmap.html:7: Fixed plurals remain: the backlog aside renders '1 change sets', and the change set page renders '1 runs'.
  - round 3 (cycle 3): 1 finding, 0 confirmed
    - f1 nit at internal/web/templates/change.html on web-change-set-page: The Runs panel reads '1 runs' for a single run (seen on /changes/01M6PFH625P3BYQAB8P6J38BBT); the backlog aside on /roadmap has the same '1 change sets' plural.
- scope (step review/review): passed after 1 cycle
  - cycle 1: passed, check session add36ab3-72bd-4ea9-859d-01a1ae7014e2 pass (said done)
  - round 1 (cycle 1): no findings
- model (step review/review): passed after 2 cycles
  - cycle 1: failed, check session c7cf927d-455e-4171-9486-01a1ae705fa8 pass (said done), critic session 71004daf-89f1-4969-957f-01a1ae7a29fc pass (said done)
  - cycle 2: passed, check session 949a92c9-23b7-46de-90ab-01a1ae8b524b pass (said done)
  - round 1 (cycle 1): 2 findings, 1 confirmed
    - f1 major at internal/web/roadmap.go:128 on web-roadmap-page: The leaf's statement says each roadmap row shows the change set's short ID, title and state, but the code shows no ID (rightly, under constraint owner-surfaces-no-ids, which postdates the leaf) and shows the change set's full intent, not its title (Row.Title is c.Intent; RoadmapChangeSet has no title field). The model and code disagree: amend the statement to drop the short ID, and either show the title (add it to the roadmap response) or say the row shows the intent. — confirmed: web-roadmap-page's statement says each row shows the change set's short ID, title and state, but internal/web/roadmap.go:128 builds Row{Title: c.Intent} with no ID (correct under owner-surfaces-no-ids, which forbids change set IDs on the web UI), and RoadmapChangeSet (internal/api/plan.go:150) has no title field, only intent. The model and the code disagree, so the statement needs amending (drop the short ID; say intent, or add a title).
    - f2 minor at internal/web/roadmap.go:105 on web-roadmap-page: The roadmap operation also returns umbrella groups (kind umbrella), but the page files every non-backlog group under Milestones, titled by its intent. The leaf only describes milestone groups and the backlog, so umbrella groups are rendered as milestones with no model statement covering them.
  - round 2 (cycle 2): no findings

## Code checks

- step review/code: pass

## Sign-offs

- signoff (visit 1): approved by rule autonomous

## Sessions

- build/build, session 56246e56-e26e-4339-b25f-01a1ae5c4f31: tokens input 50, output 15817, cache creation 80768, cache read 2140012; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session d404ffa3-3403-449c-87fe-01a1ae5ff5d3: tokens input 18, output 3090, cache creation 34962, cache read 496189; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session fb8f6edf-65a6-42d4-8cad-01a1ae603308: tokens input 48, output 15263, cache creation 83759, cache read 2067223; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session b57f3891-c212-43b2-9ca3-01a1ae63b03c: tokens input 28, output 4326, cache creation 42914, cache read 856686; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 4bd6a5fd-37ec-4b6c-b08b-01a1ae6f3fd7: tokens input 14, output 2373, cache creation 45185, cache read 411043; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 579cde3b-c2bc-4899-be0d-01a1ae6f8223: tokens input 28, output 4982, cache creation 54150, cache read 992077; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 3eb24189-72b9-46d3-b820-01a1ae6fc682: tokens input 70, output 18068, cache creation 70867, cache read 2786968; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session add36ab3-72bd-4ea9-859d-01a1ae7014e2: tokens input 16, output 1907, cache creation 42529, cache read 468314; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session c7cf927d-455e-4171-9486-01a1ae705fa8: tokens input 22, output 2863, cache creation 50994, cache read 718789; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session c4354d31-5bf4-4dbf-a6ca-01a1ae79a339: tokens input 12, output 1743, cache creation 25545, cache read 284004; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 7452807b-bb05-438d-8154-01a1ae79df3a: tokens input 24, output 3522, cache creation 37700, cache read 652918; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 71004daf-89f1-4969-957f-01a1ae7a29fc: tokens input 14, output 1421, cache creation 24081, cache read 332565; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session b85c4d10-19a0-47f0-b9de-01a1ae7bda4f: tokens input 26, output 5264, cache creation 37278, cache read 753614; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 65b4429b-e60e-4383-91f6-01a1ae8273de: tokens input 18, output 3540, cache creation 37099, cache read 503560; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 42495fce-1f7f-4e07-9e27-01a1ae82bf7c: tokens input 28, output 4988, cache creation 38874, cache read 820882; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 31bbabd7-386c-476a-97b7-01a1ae86e0fa: tokens input 16, output 2471, cache creation 34763, cache read 428128; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 00b4794a-7f2c-4265-81f2-01a1ae8a65a8: tokens input 26, output 4376, cache creation 49921, cache read 853316; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 0ebbd179-5ce0-4894-bf2d-01a1ae8ad2ff: tokens input 56, output 12841, cache creation 59893, cache read 2033767; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 949a92c9-23b7-46de-90ab-01a1ae8b524b: tokens input 18, output 2724, cache creation 46399, cache read 524756; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session e4d9a306-38c8-4de7-a63f-01a1ae91609c: tokens input 22, output 2728, cache creation 29845, cache read 569579; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 85706367-f4e2-40e6-98f2-01a1ae919d0c: tokens input 12, output 2022, cache creation 30846, cache read 309433; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 0b0d7fe8-28ec-444e-8399-01a1ae92907b: tokens input 12, output 2414, cache creation 31319, cache read 308621; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 49cfd641-2f52-4285-b767-01a1ae9457dd: tokens input 14, output 2684, cache creation 35592, cache read 373284; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 6d941432-2e4b-476e-a8ef-01a1ae95defb: tokens input 16, output 2063, cache creation 44218, cache read 466706; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 4b0924b9-1453-4e93-932e-01a1ae964e12: tokens input 54, output 11049, cache creation 53837, cache read 1874284; helper tokens input 0, output 0, cache creation 0, cache read 0

Totals: tokens input 662, output 134539, cache creation 1123338, cache read 22026718; helper tokens input 0, output 0, cache creation 0, cache read 0