samples: FactoryDemo.Api.CSharp — JSON API companion, parity with F# sibling

Zeta.sln

@@ -29,6 +29,8 @@ Project("{F2A71F9B-5D33-465A-A702-920D77279786}") = "CrmSample", "samples\CrmSam
 EndProject
 Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "src", "src", "{827E0CD3-B72D-47B6-A68D-7590B98EB39B}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "FactoryDemo.Api.CSharp", "samples\FactoryDemo.Api.CSharp\FactoryDemo.Api.CSharp.csproj", "{E1F00C7B-E03C-4DA4-8B47-E872973529D1}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -183,11 +185,25 @@ Global
 		{D44AB9CA-F491-41F4-96CE-B061238F3D6E}.Release|x64.Build.0 = Release|Any CPU
 		{D44AB9CA-F491-41F4-96CE-B061238F3D6E}.Release|x86.ActiveCfg = Release|Any CPU
 		{D44AB9CA-F491-41F4-96CE-B061238F3D6E}.Release|x86.Build.0 = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|x64.ActiveCfg = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|x64.Build.0 = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|x86.ActiveCfg = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Debug|x86.Build.0 = Debug|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|Any CPU.Build.0 = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|x64.ActiveCfg = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|x64.Build.0 = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|x86.ActiveCfg = Release|Any CPU
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1}.Release|x86.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
 	EndGlobalSection
 	GlobalSection(NestedProjects) = preSolution
 		{D44AB9CA-F491-41F4-96CE-B061238F3D6E} = {5D20AA90-6969-D8BD-9DCD-8634F4692FDA}
+		{40534D09-439E-4E5F-9A69-A73844DB674D} = {5D20AA90-6969-D8BD-9DCD-8634F4692FDA}
+		{E1F00C7B-E03C-4DA4-8B47-E872973529D1} = {5D20AA90-6969-D8BD-9DCD-8634F4692FDA}
 	EndGlobalSection
 EndGlobal

docs/BACKLOG.md

@@ -9542,6 +9542,51 @@ Aarav.
 
 ---
 
+## P2 — FactoryDemo C# sample — deterministic smoke-test startup
+
+- [ ] **Smoke-test port allocation — replace random-in-range
+  with OS-assigned ephemeral port.** Codex P2 on PR #147:
+  `samples/FactoryDemo.Api.CSharp/smoke-test.sh` currently
+  picks a port from `5100-5499` via `RANDOM`, assumes it is
+  free, and fails with an address-in-use error if the port
+  is already occupied (parallel CI jobs, local services,
+  another smoke run). This creates flaky false negatives
+  unrelated to API correctness. **Scope:** two viable
+  approaches — (a) bind `--urls "http://127.0.0.1:0"` and
+  parse the chosen port from the Kestrel startup line in
+  the log file, or (b) pre-probe ports in the range with
+  `nc -z` / `/dev/tcp` and retry until one is free.
+  Approach (a) is preferred (truly deterministic; no race
+  window between probe and bind). Sibling
+  `samples/FactoryDemo.Api.FSharp/smoke-test.sh` has the
+  same issue and should be fixed in parallel. **Effort:**
+  S (shell-only, maybe 20 lines across both scripts plus a
+  log-line parser). **Owner:** devops-engineer. **Source:**
+  Codex thread `PRRT_kwDOSF9kNM59gdjf` on PR #147.
+
+## P2 — FactoryDemo C# sample — solution project-type GUID hygiene
+
+- [ ] **Align `FactoryDemo.Api.CSharp` project-type GUID
+  with other SDK-style C# projects.** Copilot finding on
+  PR #147: `Zeta.sln` lists `FactoryDemo.Api.CSharp` under
+  the legacy C# project-type GUID
+  `{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}` while every
+  other SDK-style C# project (`Core.CSharp`,
+  `Tests.CSharp`, `Core.CSharp.Tests`) uses the modern
+  SDK-style GUID `{9A19103F-16F7-4668-BE54-9A1E7A4F7556}`.
+  Mixing GUIDs can confuse solution-filter tooling and
+  Visual-Studio solution-node grouping; the `.csproj`
+  itself is already `<Project Sdk="Microsoft.NET.Sdk.Web">`
+  so the legacy GUID is purely cosmetic drift from the
+  initial `dotnet new` template. **Scope:** change the
+  single GUID in `Zeta.sln` line 32; verify `dotnet
+  build -c Release` stays at 0W/0E. **Effort:** XS
+  (single-line GUID swap). **Owner:** devops-engineer or
+  msbuild-expert. **Source:** Copilot thread
+  `PRRT_kwDOSF9kNM59geKB` on PR #147.
+
+---
+
 ## Source of this backlog
 
 - `docs/MISSED-ITEMS-AUDIT.md` — per-round sweep

docs/hygiene-history/live-lock-audit-history.md

@@ -37,3 +37,61 @@ The full memory context is
 | date (UTC) | window | EXT | INTL | SPEC | OTHR | smell? | notes |
 |---|---:|---:|---:|---:|---:|---|---|
 | 2026-04-23 | 25 | 0% | 72% | 16% | 12% | **FIRING** | Inaugural run. Last 25 merged commits on `origin/main` contain zero src/tests/samples/bench changes. Factory has been running purely on tick-history + BACKLOG + research output for weeks. Response arc: PR #141 (ServiceTitan CRM demo sample) is the pattern-breaker; once #141 merges, the next audit should show non-zero EXT. Audit script landed this run. |
+
+## Lessons integrated
+
+Per `memory/feedback_lesson_permanence_is_how_we_beat_arc3_and_dora_2026_04_23.md`,
+every live-lock firing files a lesson here. Each lesson names the
+**signature** (what pattern preceded the smell), the **mechanism**
+(what caused it), and the **prevention** (what decisions avoid
+re-occurrence). Consult this section before opening a speculative arc
+— prevention is upstream of detection.
+
+### 2026-04-23 — tick-history-and-BACKLOG-dominance-with-zero-src
+
+- **Signature.** 25 consecutive merged commits on `origin/main` with
+  exactly zero changes to `src/`, `tests/`, `samples/`, or `bench/`.
+  Every commit was either a tick-history row, a BACKLOG row, a
+  research doc, a capability-map, or an ADR. No external-observable
+  product motion.
+- **Mechanism.** The autonomous-loop cron fires every minute per
+  `docs/AUTONOMOUS-LOOP.md`. The standing never-idle discipline
+  (`memory/feedback_never_idle_speculative_work_over_waiting.md`)
+  says speculative work is valid non-idle. But there was no
+  counter-balancing force pulling the loop toward external-code
+  work. Every tick, the lowest-friction move was another
+  tick-history append or BACKLOG grooming. Compounded over dozens of
+  ticks, the factory drifted into pure meta-work without any agent
+  catching the drift.
+- **Prevention (decisions to embed forward).**
+  1. **External-priority stack is authoritative, agent-reorderable
+     only for internal priorities.**
+     `memory/project_aaron_external_priority_stack_and_live_lock_smell_2026_04_23.md`
+     names Aaron's stack (ServiceTitan+UI / Aurora / multi-algebra
+     DB / cutting-edge persistence) as externally-set; the agent
+     owns internal priorities but not the external stack's
+     ordering. Speculative work lives under internal;
+     external-priority work takes precedence when the ratio tilts.
+  2. **Live-lock audit runs at round-close as a gate-not-a-report.**
+     `tools/audit/live-lock-audit.sh` exits 1 when EXT < 20%;
+     round-close should check this signal. A smell-firing
+     round-close must include at least one external-priority
+     increment in the next round's plan before the close ledger
+     accepts.
+  3. **Speculative-work permit requires external-ratio check.**
+     Before opening a new speculative arc (research doc, large
+     BACKLOG row, capability map), agent reads the current audit
+     ratio. If smell is firing, no new speculative arcs open until
+     one external-priority increment lands. This is an *agent-
+     internal discipline*, not a blocking rule — but it gets cited
+     in the commit message ("audit EXT=X%, smell=not-firing,
+     speculative arc opens") so the discipline is visible.
+  4. **Tick-history rows are NOT external work.** The tick-history
+     append is ledger-keeping, not product motion. Counting it as
+     "forward motion" was the silent-drift mechanism. Agents
+     should explicitly describe tick-history work as INTL and
+     pair it with EXT work in the same tick when the smell is
+     near firing.
+- **Open carry-forward.** The round-close-ladder wiring is a
+  follow-up — the audit script is landed, but it is not yet
+  invoked automatically at round-close. BACKLOG P1 row filed.