fix(detection): support list-format JSON output from Claude Code CLI …

[yazeedallabadi24]

What type of PR is this? (check all applicable)

• Refactor
• Feature
• Bug Fix
• Optimization
• Documentation Update

Related issue:
N/A

What changed?
Updated _invoke_claude_reasoning and _extract_mcp_tool_usage in Detection/guardrail/adr_agent/adr_baseline.py to handle list/array formatted JSON outputs from Claude Code CLI. When --output-format json yields an event array, the parser extracts the entry with type: result (falling back to the last element if absent).

Why?
Modern Claude Code CLI versions (2.1+, e.g. v2.1.274) output a stream of JSON events formatted as an array [{"type": "system", ...}, ..., {"type": "result", ...}] rather than a single top-level dictionary. Previously, json.loads(stdout) produced a list, triggering an unhandled 'list' object has no attribute 'get' exception.

How did you test it?

1. Live CLI Verification: Tested against Claude Code CLI v2.1.274, confirming both _invoke_claude_reasoning and _extract_mcp_tool_usage successfully parse the event list to extract reasoning verdicts and session IDs.
2. Backward Compatibility: Verified that standard dictionary outputs (from earlier CLI versions) remain unaffected.
3. Detection Test Suite: Ran uv run pytest tests/ in Detection/ — all 165 tests passed (including all 63 in test_adr_baseline.py).
4. Sensor Test Suite: Ran uv run pytest tests/ in Sensor/ — all 287 tests passed.

Potential risks
None. Single-dict responses bypass the list handler and execute existing behavior unchanged.

[CLAassistant]

All committers have signed the CLA.

[barisozbas]

LGTM!