What problem does it solve?
After a refresh, checkout switch, or preset/MCP change, developers often cannot tell whether the lingtai agent is actually running the new code — merged fixes appear to fail because stale imports, polluted PYTHONPATH, or unrebuilt long-lived runtime objects silently persist. This Skill provides a read-only diagnostic checklist to prove what code is live and report it safely.
Core Features & Use Cases
- Runtime source probing: Identify which lingtai package the agent venv imports, whether it is an editable install, and the git HEAD behind it, including PYTHONPATH pollution detection.
- Binary and symlink verification: Confirm the active lingtai-tui/portal binaries resolve to the intended worktree and that rebuilds actually landed on PATH.
- Live object lifecycle checks: Verify that long-lived services, adapters, and caches were genuinely rebuilt after refresh, not just that new source was imported.
- Safe evidence reporting: Produce a compact, redacted evidence pack with secrets, tokens, chat IDs, and private paths removed.
- Use Case: A merged kernel fix shows no effect after refresh — run the self-check to discover the agent imports a different checkout, fast-forward it, refresh, and confirm the fix via an in-situ probe.
Quick Start
Run the runtime self-check to verify which lingtai source, binary, and MCP modules this agent is actually executing and report the evidence with secrets redacted.
