test(mcp): isolate the invalid-action trace test from on-disk state

[basil-k-aji-dev]

Closes #62.

Problem

test_mobile_inspect_trace_invalid_action does not request the temp_trace_env fixture the rest of the module uses, so it runs against whatever trace_store.TRACES_DIR resolves to. mobile_inspect_trace looks for data_engine.db there, falls back to <project_root>/traces/data_engine.db, and returns "Database not found" before it validates action at all — so the unsupported-action branch the test names is never reached either way.

The result therefore reports leftover disk state, not the logic under test.

Reproduction on this host

The project-root database exists only because running the suite creates it, which is exactly what makes this self-concealing — it passes on the second run and fails on a clean checkout:

$ ls traces/data_engine.db
traces/data_engine.db
$ uv run pytest -q tests/unit/mcp/test_mcp_tools.py::test_mobile_inspect_trace_invalid_action
1 passed

$ mv traces /tmp/traces_bak          # simulate a clean checkout
$ uv run pytest -q tests/unit/mcp/test_mcp_tools.py::test_mobile_inspect_trace_invalid_action
AssertionError: assert 'not supported' in 'Data engine database does not exist at .../traces/data_engine.db yet.'
1 failed

I hit this accidentally mid-session: a full-suite run overlapped with the window in which I had moved traces/ aside, and this single test flipped to failing in that run and back to passing afterwards, with no code change in between.

Change

Give the test the same isolated trace directory the other tests in the module use, and seed an empty data_engine.db in it. An unknown action never reads the database — the only gate before the final else is os.path.exists(db_path) — so an empty file is enough to get past it and into the branch under test. Seeding the temp database also stops the <project_root>/traces fallback from being consulted at all, which is what made temp_trace_env alone insufficient here.

Also assert that the offending action name is echoed back, since that is the part of the message a caller acts on.

Scope note

I deliberately did not move action validation ahead of the database check in mcp_server/tools/inspect_trace.py. It would arguably be a better error for an unknown action with no database present, but it changes the tool's response for a real caller and is a separate decision from making this test honest. Happy to follow up if you'd prefer that.

Verification

The decisive property is that the outcome no longer depends on the project-root database:

	traces/data_engine.db present	absent
this branch	passed	passed
main	passed	failed

Command	Result
uv run pytest -q tests/unit/mcp/test_mcp_tools.py	25 passed
uv run pytest -q (full deterministic suite)	2096 passed, same 89 pre-existing failures as main — 0 new
uv run ruff check / ruff format --check on the touched file	clean

make typecheck could not be run in my environment — the pyright wheel downloads a Node runtime on first use and my sandbox blocks it. The touched file is not in pyright-core.json.

[RohanOnKeys]

close this pr, a solution is already on way (#66 .)
please read thru the codebase before you commit anything

[basil-k-aji-dev]

Closing as a duplicate. #66 by @RohanOnKeys already covers issue #62 and was opened first — that one should get the review, not this.

My mistake: I didn't check the open PR queue for an existing claim before sending this. Sorry for the noise.