@@ -0,0 +1,26 @@
+---
+description: All project texts (prompts, guardrails, skills, rules, markers, READMEs, fixtures) are written in English
+alwaysApply: true
+---
+
+# English-only texts
+
+Everything the model or a developer reads in this repo is English:
+
+- system prompts, nudges, guardrail/oracle notes, tool results and markers (`…[clipped: …]`, `(already in context …)`)
+- skills and rules (`extensions/codeforge/resources/**`, `governance/builtins.ts`)
+- status/activity labels, reports (`BUGS.md`, plan report), READMEs, `HEURISTICS.md`
+- benchmark prompts (`task.json`) and fixture docs/cards (`Depends on:`, `Acceptance criteria`)
+
+Portuguese is allowed only in **matchers of user input** (regexes, skill `triggers`, stopwords), because users write in Portuguese:
+
+```ts
+// ✅ matcher — keeps recognizing Portuguese requests
+/\b(review|revis\w*|encontr\w* (os )?bugs?)\b/i
+// ❌ text sent to the model
+`…[cortado: mostradas linhas 1-${n}]`
+// ✅
+`…[clipped: showing lines 1-${n} of ${total}]`
+```
+
+Chat replies to the user follow the user's language; this rule is about files in the repo.
@@ -83,3 +83,5 @@ vscode/
 
 # Local planning docs (keep offline)
 BLUEPRINT-02-Open-Source-AI-IDE-Revised-Embedded-AI-Platform.md
+
+benchmarks/results/
@@ -0,0 +1,48 @@
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio Version 17
+VisualStudioVersion = 17.5.2.0
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "benchmarks", "benchmarks", "{66320409-64EC-F7C5-3DEF-65E7510DAAD1}"
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "fixtures", "fixtures", "{AFA72641-1A30-5FD9-46FF-8D3CF9AF32B9}"
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "billing", "billing", "{D9734131-DA50-2FA4-E292-8A411DE028FF}"
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "src", "src", "{87F2DFEF-6771-6C5F-308A-534D1B05A8F9}"
+EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Acme.Billing", "benchmarks\fixtures\billing\src\Acme.Billing\Acme.Billing.csproj", "{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867}"
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "tests", "tests", "{75E704B8-F95A-BA8B-0982-342EDA781354}"
+EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "Acme.Billing.Tests", "benchmarks\fixtures\billing\tests\Acme.Billing.Tests\Acme.Billing.Tests.csproj", "{2CBCF198-0448-A543-E33E-B0A55F3610E8}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|Any CPU = Debug|Any CPU
+		Release|Any CPU = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867}.Release|Any CPU.Build.0 = Release|Any CPU
+		{2CBCF198-0448-A543-E33E-B0A55F3610E8}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{2CBCF198-0448-A543-E33E-B0A55F3610E8}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{2CBCF198-0448-A543-E33E-B0A55F3610E8}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{2CBCF198-0448-A543-E33E-B0A55F3610E8}.Release|Any CPU.Build.0 = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+	GlobalSection(NestedProjects) = preSolution
+		{AFA72641-1A30-5FD9-46FF-8D3CF9AF32B9} = {66320409-64EC-F7C5-3DEF-65E7510DAAD1}
+		{D9734131-DA50-2FA4-E292-8A411DE028FF} = {AFA72641-1A30-5FD9-46FF-8D3CF9AF32B9}
+		{87F2DFEF-6771-6C5F-308A-534D1B05A8F9} = {D9734131-DA50-2FA4-E292-8A411DE028FF}
+		{F2A9ADC8-9C72-F8A6-E9F4-2AE26A42D867} = {87F2DFEF-6771-6C5F-308A-534D1B05A8F9}
+		{75E704B8-F95A-BA8B-0982-342EDA781354} = {D9734131-DA50-2FA4-E292-8A411DE028FF}
+		{2CBCF198-0448-A543-E33E-B0A55F3610E8} = {75E704B8-F95A-BA8B-0982-342EDA781354}
+	EndGlobalSection
+	GlobalSection(ExtensibilityGlobals) = postSolution
+		SolutionGuid = {DC61E0B0-E006-49BD-AAAE-6F2B47245DB7}
+	EndGlobalSection
+EndGlobal
@@ -1,6 +1,38 @@
 # CodeForge Benchmarks
 
-This directory contains benchmark suites for evaluating CodeForge's AI capabilities.
+## Harness benchmark (available now)
+
+Semi-automatic suite: the agent runs inside CodeForge; `check` scores with hidden tests + session metrics.
+
+### JavaScript (`fixtures/shop`) — T1–T4
+Quick harness smoke (node). Good for catching retrieve/read loops; **not** your real C# workload.
+
+### C# / .NET 10 (`fixtures/billing` + `scaffold`) — CS1–CS4
+Closer to ContextRouter: `dotnet test`, classlibs, docs-driven features, VIP navigation, and a **scaffold** task (create solution from a C01 card).
+
+```bash
+npm run bench -- list
+npm run bench -- setup CS1 --label bonsai --open   # or CS2 / CS3 / CS4
+# New Chat in that folder → bonsai → paste the printed prompt
+npm run bench -- check latest
+npm run bench -- report
+npm run bench -- selftest CS                       # validates fixtures + solutions
+```
+
+Score (0–100): hidden 60 · visible 15 · protected untouched 10 · checks 10 · no junk files 5.
+
+### Long and review (`fixtures/router-cards`, `fixtures/billing-bugs`) — CS5–CS6
+
+- **CS5 (review)**: 6 bugs planted in `billing-bugs` (visible tests still pass). The agent writes
+  `BUGS.md` with `path:line` references; the manifest lives in `tasks/CS5-review-billing-bugs/bugs.manifest.json`
+  (outside the fixture). Score: F1 × 80 · protected 10 · `BUGS.md` exists 10 (±3 line tolerance).
+- **CS6 (long)**: cards C01–C05 with "Depends on:" lines. Hidden tests run **per card** (only that
+  card's tests are compiled in); the hidden ratio is the average over cards.
+
+Every result also reports `falseDone` (the agent claimed green while tests were red) and
+`harness` (chars of harness text injected into the conversation).
+
+## Planned suites
 
 ## Categories