Daily tech-debt sweep — 2026-09-22

.codex/TECH_DEBT_QUEUE.md

@@ -33,7 +33,7 @@ violation, never edit a baseline the code still triggers.
   `src/Sections/Humans.<Section>/Docs/<Section>.md` + `Docs/data-access.md`.
   Debt is whatever diverges from that model; find it by comparing a section against
   the model, not by consuming a frozen list.
-- **Surface / interconnectivity baseline:** `dotnet build Humans.slnx -v quiet`, then
+- **Surface / interconnectivity baseline:** `dotnet build Humans.slnx -v quiet -clp:ErrorsOnly`, then
   `reforge surface-score --all --top-symbols 200 --format Json` (score a **built**
   solution — unbuilt under-reports ~4%). Rank sections by the Section Refactor History
   table in `docs/architecture/maintenance-log.md`. The score is a detector, not an
@@ -71,8 +71,8 @@ violation, never edit a baseline the code still triggers.
    *Current state* below **by rewriting it**.
 2. Pick one item by the priority order. Write a one-sentence architecture thesis; if
    the thesis is "a number goes down", pick something else.
-3. Make the smallest real refactor. Targeted section tests + `dotnet build Humans.slnx -v quiet`
-   per change; full `dotnet test Humans.slnx -v quiet` before any push.
+3. Make the smallest real refactor. Targeted section tests + `dotnet build Humans.slnx -v quiet -clp:ErrorsOnly`
+   per change; full `dotnet test Humans.slnx -v quiet -clp:ErrorsOnly` before any push.
 4. One coherent improvement per commit; push the branch; open/refresh the PR
    (`memory/process/always-open-a-pr.md`).
 5. When stopping: rewrite *Current state* (including *Needs Peter*), leave the

.codex/bug-hunt-prompt.md

@@ -76,7 +76,7 @@ Do not modify:
 ## Build Command
 
 ```
-dotnet build Humans.slnx -v q && dotnet test Humans.slnx -v q --filter "FullyQualifiedName~Application"
+dotnet build Humans.slnx -v q -clp:ErrorsOnly && dotnet test Humans.slnx -v q -clp:ErrorsOnly --filter "FullyQualifiedName~Application"
 ```
 
 ## Phase 1: Razor Rendering & HTML Structure *(highest frequency — 15+ historical fixes)*

.codex/skills/humans-bug-hunt/SKILL.md

@@ -27,8 +27,8 @@ Run recurring autonomous bug-hunt passes in this repository.
 2. Search for one high-confidence bug at a time, using repo patterns rather than a fixed checklist.
 3. Implement the smallest defensible fix.
 4. Add or extend tests when practical.
-5. Run targeted verification, plus `dotnet build Humans.slnx --disable-build-servers -v q`.
-6. Periodically run `dotnet test Humans.slnx --no-build --disable-build-servers -v q --filter "FullyQualifiedName~Application"`.
+5. Run targeted verification, plus `dotnet build Humans.slnx --disable-build-servers -v q -clp:ErrorsOnly`.
+6. Periodically run `dotnet test Humans.slnx --no-build --disable-build-servers -v q -clp:ErrorsOnly --filter "FullyQualifiedName~Application"`.
 7. Commit each fix separately and push the branch after verified progress.
 8. Continue until remaining ideas are speculative or require forbidden areas.
 

.codex/skills/humans-refactor/SKILL.md

@@ -161,7 +161,7 @@ Then repeat until stasis:
 3. Pick the highest-leverage cohesive improvement, not just the highest scoring rule.
 4. Write a candidate thesis in the run notes: what concept will be deleted, which responsibility moves to its rightful owner, or which duplicated/cross-section path disappears. If the thesis is just "the score drops", reject the candidate before editing.
 5. Make the change.
-6. Run targeted tests and `dotnet build Humans.slnx --disable-build-servers -v q`.
+6. Run targeted tests and `dotnet build Humans.slnx --disable-build-servers -v q -clp:ErrorsOnly`.
 7. Run Reforge after the change.
 8. Run the score-blind architecture-review gate.
 9. If accepted, commit and push. If rework/reject, improve or abandon before committing.

.codex/skills/humans-tech-debt/SKILL.md

@@ -30,9 +30,9 @@ Run recurring autonomous tech-debt reduction passes in this repository.
 3. Before editing, write a one-sentence architecture thesis: what concept will be deleted, what responsibility will move to its rightful owner, or what duplication/coupling will disappear. If the thesis is "the score drops", abandon the candidate.
 4. Make the smallest coherent improvement that reduces divergence, duplication, misplaced responsibility, or durable public surface.
 5. Add or extend tests when practical.
-6. Run targeted verification, plus `dotnet build Humans.slnx --disable-build-servers -v q`.
+6. Run targeted verification, plus `dotnet build Humans.slnx --disable-build-servers -v q -clp:ErrorsOnly`.
 7. Run a score-blind second pass before commit. Review only the diff, the architecture thesis, and verification. Reject the change if it would not be worth keeping without metric movement.
-8. Run the full `dotnet test Humans.slnx --no-build --disable-build-servers -v quiet` gate before any push.
+8. Run the full `dotnet test Humans.slnx --no-build --disable-build-servers -v quiet -clp:ErrorsOnly` gate before any push.
 9. Commit each accepted improvement separately and push the branch after verified progress.
 10. Continue until remaining ideas are low-value, speculative, blocked by forbidden areas, or only reducible through metric-gaming changes.
 

.codex/tech-debt-prompt.md

@@ -32,7 +32,7 @@ file's *Current state* instead of restarting discovery.
   `docs/sections/SECTION-TEMPLATE.md`, `docs/architecture/design-rules.md`, each
   section's `Docs/<Section>.md` + `Docs/data-access.md`. Divergence from the model is
   the debt.
-- Surface / coupling baseline: `dotnet build Humans.slnx -v quiet`, then
+- Surface / coupling baseline: `dotnet build Humans.slnx -v quiet -clp:ErrorsOnly`, then
   `reforge surface-score --all` (built solution only). Reduce it through
   architecturally-real deletions; the score is a detector, never the objective.
 
@@ -46,7 +46,7 @@ file's *Current state* instead of restarting discovery.
   `local/tech-debt-runs/<run-id>/` either way.
 - One coherent improvement per commit, each with a one-sentence architecture thesis
   that stands without score movement. Targeted section tests + build per change;
-  full `dotnet test Humans.slnx -v quiet` before any push. `-v quiet` always.
+  full `dotnet test Humans.slnx -v quiet -clp:ErrorsOnly` before any push. `-v quiet` always.
 - Honor every boundary in the queue file's *Boundaries* section — especially: debt
   only (never feature follow-ups, even fully-specced ones), no authorization/privacy
   shape changes, no reverting documented test-infrastructure decisions, and new public surface goes to *Needs Peter* instead of into the code.

docs/architecture/debt-ledger.yml

@@ -395,10 +395,6 @@ inbox:
   # sweep scope entirely); PasswordGenerator CSPRNG (explicitly deferred by
   # Peter, 2026-04-24); BudgetRepository ResponsibleTeam .Include (covered by
   # grandfathered-hum0024-nav-strip).
-  - added: 2026-08-11
-    id: CENTRAL-14
-    what: "EventSettingsConfiguration.cs:44-47 declares HasDefaultValue(-25/-16/-9/-4) on four int offsets without HasSentinel — EF's int sentinel stays 0, so explicitly assigning 0 to any of these offsets is skipped and the DB default written instead (int-flavoured bool-sentinel trap). Pre-existing; surfaced by the EF migration review of the Shifts peel (peel 13, nobodies-collective/Humans#858)"
-    review: light
   - added: 2026-08-07
     id: CENTRAL-12
     what: "Shifts section's own Web layer (ShiftAdminController, ShiftsController, VolunteerTrackingController, ShiftViewModels.cs, ShiftAdminPageBuilder, ShiftBrowsePageBuilder, ShiftDashboardPageBuilder, ShiftBrowseMapper, ShiftFilterResolver, ShiftSignupBucketer, ShiftVolunteerSearchBuilder, DevelopmentDashboardSeeder, ShiftSignupsViewComponent) carries the EF EventSettings entity as a view-model/builder field on read-only display paths — should be BurnSettingsInfo per nobodies-collective/Humans#809's acceptance criteria. Distinct from the write/edit path (EventSettingsFormMapper), which legitimately keeps the entity. ~10 interconnected files forming one section's own display pipeline — warrants its own dedicated section-scoped PR rather than folding into #809's cross-section batch (found while working nobodies-collective/Humans#809)"
@@ -583,10 +579,6 @@ inbox:
     id: CENTRAL-59
     what: "src/Sections/Humans.Users.Contracts/IUserInfoInvalidator.cs:24 deliberately omits the IInvalidator marker (its own comment says so, debt nobodies-collective/Humans#805), so HUM0028 cannot see it and it needs no [Grandfathered] marker — one invalidator dodging the ratchet by being invisible to it. Marking it (or recording the exemption where the analyzer reads it) is small and separable from the big invalidator rework the HUM0028 theme is parked on. Found by the 2026-09-20 accepted-smell audit."
     review: light
-  - added: 2026-09-20
-    id: CENTRAL-60
-    what: "The pre-existing Codex prompt/skill files run dotnet with `-v quiet`/`-v q` but without the `-clp:ErrorsOnly` that memory/process/dotnet-verbosity-quiet.md requires (.codex/tech-debt-prompt.md, .codex/bug-hunt-prompt.md, .codex/TECH_DEBT_QUEUE.md, .codex/skills/humans-{tech-debt,refactor,bug-hunt}/SKILL.md). Without it the obsolete-warning wall floods the context of every Codex run that follows them, which costs most on a time-budgeted unattended run. Mechanical fix, no judgement needed. Found 2026-09-20 while adding the flag to the nightly debt runner in peterdrier/Humans#1759, which deliberately did not widen to these files."
-    review: light
   - added: 2026-09-20
     id: CENTRAL-61
     what: "Two pre-existing Codex runner scripts use `git -C <path>`, which memory/process/never-use-git-dash-c.md marks a HARD RULE with no exceptions: .codex/run-weekly-bug-hunt.sh and .codex/cleanup-merged-bug-hunt-worktrees.sh. Fix is mechanical — run each group in a `(cd \"$dir\" && git ...)` subshell, as .codex/cron/run-daily-debt.sh now does. Worth noting the atom's stated rationale (a wrong shell folder; defeating Bash allowlist prefixes) is about agent-issued commands, so Peter may want to say whether the rule binds committed scripts at all rather than have it re-found every audit. Found 2026-09-20 while fixing the same violation in the nightly debt runner in peterdrier/Humans#1759."

src/Sections/Humans.Camps/Data/CampRepository.Roles.cs

@@ -135,7 +135,6 @@ public async Task<IReadOnlyList<CampRoleAssignment>> GetAssignmentsForSeasonAsyn
             .Include(a => a.Definition)
             .Include(a => a.CampMember)
             .Where(a => a.CampSeasonId == campSeasonId)
-            .OrderBy(a => a.Definition.SortOrder).ThenBy(a => a.AssignedAt)
             .ToListAsync(ct);
     }
 

src/Sections/Humans.Email/Docs/debt.yml

@@ -2,8 +2,4 @@
 # /debt-sweep pools every src/Sections/*/Docs/debt.yml into the same inbox as the central ledger.
 version: 1
 next_id: 3
-inbox:
-  - added: 2026-09-11
-    id: EMAIL-2
-    what: "src/Sections/Humans.Email/Services/EmailRenderer.cs HTML-encodes these subject lines: Email_IssueComment_Subject, Email_CoordinatorRotaMessage_Subject, Email_CoordinatorTeamRotasMessage_Subject, Email_GoogleGroupRemoval_LossOfAccess_Subject, Email_GoogleDriveRemoval_LossOfAccess_Subject, Email_GoogleAccessRemoval_SecondaryCleanup_Subject. A Subject header is plain text, so a title containing & or < reaches the recipient as R&amp;D. The assembly-vote subjects had the same bug and were fixed in peterdrier/Humans#1649; these are left alone as out of that PR's scope."
-    review: light
+inbox: []