netease-game-instrumentation

Use hypothesis-driven instrumentation: list 1-3 concrete hypotheses, add temporary _dbgLog calls that write NDJSON entries to a per-problem log file, trigger the logic in-game, then analyze logs grouped by hypothesisId before writing any fix.

Add _dbgLog calls at entity placement and render points recording entityPos, playerPos, and object size as structured data. Compare logged coordinates against expected values to confirm the offset direction, then apply the correction and retest.

Yes, when the mcdk_dev MCP server is available. The AI can reload the game, trigger target logic with execute_code, capture the game window, and read logs directly, eliminating the wait for manual user testing.

NDJSON writes one JSON object per line with required fields: id, timestamp, location, message, data, runId, and hypothesisId. This structure enables sorting by timestamp, grouping by hypothesis, and comparing runs across test sessions.

Use execute_code one-off probes when hypotheses are simple (1-2) and do not involve timing. Use persistent source-file instrumentation when you need continuous logging across ticks, callbacks, multiple modules, or a reproducible evidence chain.

Temporary _dbgLog functions, imports, and log files pollute the workspace and can affect production behavior. The workflow mandates deleting all DEVONLY-marked blocks, debug imports, notification calls, and debug_*.log files after the fix is verified.