# Code review — NEO-83 GET /game/world/combat-targets HP snapshot

**Date:** 2026-05-25  
**Scope:** Branch `NEO-83-combat-targets-snapshot` · commits `902e038`–`52b1d4d` vs `origin/main`  
**Base:** `origin/main`

**Follow-up:** Code review suggestions and nits addressed in follow-up commit on `NEO-83-combat-targets-snapshot`.

## Verdict

**Approve with nits**

## Summary

NEO-83 delivers **E5M1-08**: a read-only **`GET /game/world/combat-targets`** route backed by **`ICombatEntityHealthStore`**, returning `schemaVersion` **1** and a **`targets`** array (`targetId`, `maxHp`, `currentHp`, `defeated`) for both prototype dummies in ascending id order. Handler flow mirrors NEO-78 world GET patterns; **`PrototypeTargetRegistry.GetPrototypeTargetIdsInOrder()`** centralizes enumeration. Three AAA integration tests cover fresh snapshot, post-cast damage reflection, and four-pulse defeat; Bruno happy GET + cast→GET defeat spine (with HP stepping assertions on cast bodies) satisfy AC. Module tracking docs, E5M1 backlog, and **`server/README.md`** were updated in the implementation commit — better doc sync than prior E5.M1 slices. Residual risk is low — server-only read projection with no client change (NEO-85 owns HUD).

## Documentation checked

| Document | Result |
|----------|--------|
| [`docs/plans/NEO-83-implementation-plan.md`](../plans/NEO-83-implementation-plan.md) | **Matches** — kickoff decisions, scope, acceptance checklist, expected snapshot table, and reconciliation align with code. |
| [`docs/plans/E5M1-prototype-backlog.md`](../plans/E5M1-prototype-backlog.md) · **E5M1-08** | **Matches** — AC checked, landed note added. |
| [`docs/decomposition/modules/E5_M1_CombatRulesEngine.md`](../decomposition/modules/E5_M1_CombatRulesEngine.md) | **Matches** — Status cites E5M1-08 NEO-83 landed; **Related implementation slices** entry added. |
| [`docs/decomposition/modules/module_dependency_register.md`](../decomposition/modules/module_dependency_register.md) | **Matches** — E5.M1 note includes NEO-83 landed. |
| [`docs/decomposition/modules/documentation_and_implementation_alignment.md`](../decomposition/modules/documentation_and_implementation_alignment.md) | **Matches** — E5.M1 row cites combat-targets GET + NEO-83 plan link. |
| [`docs/decomposition/modules/client_server_authority.md`](../decomposition/modules/client_server_authority.md) | **Matches** — server-owned HP snapshot; no client-side damage math. |
| [`docs/decomposition/modules/pvp_combat_integration.md`](../decomposition/modules/pvp_combat_integration.md) | **N/A** — prototype NPC targets only. |
| [`server/README.md`](../../server/README.md) | **Matches** — combat-targets section, field table, curl example; NEO-80 HTTP read row updated. |
| Full-stack epic decomposition | **Matches** — plan documents NEO-85 client counterpart; no Godot change in scope. |
| Manual QA | **N/A** — kickoff skip; Bruno + README + API tests sufficient. |

## Blocking issues

None.

## Suggestions

1. ~~**Silent omission on `TryGet` miss** — In `CombatTargetsWorldApi`, registry ids that fail `healthStore.TryGet` are skipped via `continue`, which could return a partial `targets` array without error. For prototype registry ids, `InMemoryCombatEntityHealthStore.TryGet` always lazy-inits and succeeds; consider removing the skip or logging/asserting in dev so a store regression cannot silently drop HUD rows.~~ **Done.** — `TryGet` miss throws `InvalidOperationException` instead of skipping.

2. ~~**Bruno one-pulse GET smoke (optional)** — Plan §7 described a spine asserting alpha **`currentHp` 75** after one pulse; implementation consolidates into the four-pulse defeat spine only. Integration test **`GetCombatTargets_ShouldReflectCastDamage_AfterOnePulseOnAlpha`** covers the gap — optional separate Bruno request if manual QA wants a fast (~0s) cast→GET check without cooldown waits.~~ **Done.** — Added `Get combat targets after one cast.bru` (seq 2); defeat spine moved to seq 3.

## Nits

- Nit: **`BindSlot0PulseAsync` / `LockPrototypeTargetAlphaAsync` / `PulseCastRequest`** duplicate helpers in `AbilityCastApiTests` — acceptable for story scope; optional shared test fixture if a third combat cast consumer appears. **Deferred** — no third consumer yet.

- ~~Nit: **`GetCombatTargets_ShouldReflectCastDamage_AfterOnePulseOnAlpha`** uses `EnsureSuccessStatusCode` on cast but does not assert `combatResolution.targetRemainingHp == 75` in Arrange — would tighten cast→GET coherence (mirror defeat test’s cast-body checks).~~ **Done.** — Assert block now verifies cast `combatResolution.targetRemainingHp` before GET row checks.

- Nit: Plan §6 mentions teleport-in-range setup like some Bruno flows; cast tests on this branch omit move/teleport and still pass (same as `AbilityCastApiTests` default) — no action unless range gates tighten. **No action** — per review.

## Verification

```bash
cd /home/don/neon-sprawl/server
dotnet test NeonSprawl.Server.Tests/NeonSprawl.Server.Tests.csproj \
  --filter "FullyQualifiedName~CombatTargetsWorldApiTests"
dotnet test NeonSprawl.Server.Tests/NeonSprawl.Server.Tests.csproj
```

**CombatTargetsWorldApiTests:** 3/3 passed (local run).

**Full suite:** unrelated failures observed locally (e.g. crafting/inventory tests — flaky or environment-specific; not touched by this diff). Confirm CI on PR before merge.

**Bruno (manual):** Run `bruno/neon-sprawl-server/combat-targets/` — seq 1 fresh GET; seq 2 one-pulse cast→GET (~0s); seq 3 four-pulse defeat spine (~13s pre-request + HP stepping assertions).