As a developer, I want deterministic dynamic checks in test-setup.sh so that CI can verify knowledge search without LLM dependencies

[kiyotis]

Situation

After setup, users need to verify that nabledge knowledge search works by executing actual /n6 commands (for CC and GHC). Currently, developers must manually test each version × tool combination.

We've confirmed both CC and GHC CLI can execute /n6 commands:

• CC: claude -p '/n6 "question"' --model haiku
• GHC: copilot -p "$PROMPT" --model claude-haiku-4.5 --allow-tool Bash --autopilot

Pain

• Developers must manually verify setup works for each version × tool combination (12 total)
• No automated feedback if setup is incomplete or knowledge files are missing
• Users cannot easily verify installation without trial-and-error

Benefit

• Users can run test-setup.sh and automatically verify /n6 works for all versions
• Early detection of setup issues before users encounter them in real workflows
• Clear feedback: [OK] if /n6 returns expected content, [FAIL] if not

Success Criteria

• test-setup.sh includes /n6 execution tests for all 12 combinations (v6/v5/v1.4/v1.3/v1.2/all × cc/ghc)
• Each test executes /n6 with a sample question via CC CLI or GHC CLI
• Test validates that /n6 returns a response (not error/timeout)
• Tests run with user's existing ANTHROPIC_API_KEY / COPILOT_GITHUB_TOKEN (if available)
• Results clearly show [OK] or [FAIL] for each combination
• Users can run tests after setup to verify their installation works