[PATCH 39/40] CLAUDE.md: add don't BS rules

[Luis Chamberlain <mcgrof@kernel.org>]

Add prominent section explicitly forbidding AI assistance from
generating, synthesizing, mocking, or fabricating any test results of
any kind. This rule was added after catching AI attempting to create
synthetic performance data when encountering issues with VM provisioning
instead of properly debugging and running actual tests on real systems.

kdevops is a real Linux kernel testing framework where all results must
come from actual execution on real Device Under Test (DUT) systems.
Fake results undermine the entire purpose and could affect important
development decisions.

Generated-by: Claude AI
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
---
 CLAUDE.md | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/CLAUDE.md b/CLAUDE.md
index 7764bc67..c52774b5 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -13,6 +13,37 @@ and complex testing laboratories for kernel subsystems.
 **Main Repository**: https://github.com/linux-kdevops/kdevops
 **License**: copyleft-next-0.3.1
 
+## ⚠️ CRITICAL RULE: NEVER FAKE TEST RESULTS
+
+**ABSOLUTELY NEVER generate, synthesize, mock, or fake test results of any kind.**
+
+kdevops is a professional testing framework used for Linux kernel development and
+performance analysis. All test results MUST come from actual execution on real
+Device Under Test (DUT) systems.
+
+### What this means:
+- **NEVER** create synthetic performance data or benchmark results
+- **NEVER** generate mock IOPS, bandwidth, latency, or throughput numbers
+- **NEVER** create fake JSON test output or fabricated metrics
+- **ALWAYS** run actual tests on real VMs, bare metal, or cloud instances
+- **ALWAYS** collect real data from actual fio, fstests, blktests execution
+- **ALWAYS** verify that test systems are running and accessible before analysis
+
+### Valid approaches:
+- ✅ Execute real tests: `ansible -i hosts all -m shell -a "fio --output-format=json ..."`
+- ✅ Collect actual JSON output from running VMs
+- ✅ Parse and analyze real performance data from live systems
+- ✅ Generate graphs and reports from actual test execution results
+
+### Forbidden approaches:
+- ❌ Creating synthetic performance numbers for demonstrations
+- ❌ Generating mock test results for visualization examples
+- ❌ Fabricating benchmark data to show workflows
+- ❌ Using placeholder values like "let's assume 50K IOPS"
+
+**Violation of this rule undermines the entire purpose of the kdevops testing framework
+and produces misleading results that could affect important development decisions.**
+
 ## Core Architecture
 
 ### Build System
-- 
2.47.2