test: clear the audit debt in the test suite

[0xKT]

Summary

A read-only audit of all 298 test files found the suite in good shape -- exactly one genuinely fake test -- with the real debt sitting in file names and a scattering of assertions that passed for the wrong reason. This clears that, and fixes the bugs found while clearing it.

Three dead tests dealt with. Two removed because neither could ever assert anything:

• test_chat_e2e.py was a module-level skip over three NotImplementedError stubs. The wire-up it waited for has shipped (tui_commands instantiates SubscriptionEmitter) and the proposal doc it cited is gone.
• test_tui_exit_e2e.py asserted that building the agent loop leaves lancedb's thread live. It does not: the loop is constructed but never started, and the memory plugin talks to everos over HTTP, so lancedb is not even imported. The assertion could not hold in any configuration.

The third, slashParity.test.ts, shelled out to python for a raven_cli.commands.COMMAND_REGISTRY that no longer exists, and skipped all three of its cases when the import failed -- three skips on every run. The parity it wanted needs no python: createSlashHandler.ts:45-51 routes a slash locally when findSlashCommand resolves it and otherwise hands it to slash.exec, a CLI subprocess a mutation cannot reach the live session from. It now asserts that predicate directly. NATIVE_MUTATING_COMMANDS and the three-way classifyRoute went with it: all six names in that set are also in the local registry, so the native branch could not change any outcome.

19 files renamed off phase, ticket and bug codes (AGENTS.md section 5.1), including one reversed CLI name and two integration files moved onto a legal <kind>. Reference updates ride along: .pre-commit-config.yaml names one of these files in a detect-private-key exclude, and three docs pages point at them.

Assertions that passed for the wrong reason, tightened: eight over-broad pytest.raises(Exception) narrowed to the type each site actually raises; an except clause that swallowed the AssertionError its own fakes use as a tripwire; a hasattr probe replaced by binding a broker and checking it stuck; and a recall assertion that read as a pass on an empty result set.

Dead weight removed: ten unused locals, three dead code blocks, six importorskip guards on hard dependencies, two hardcoded /tmp paths, and ticket codes left in docstrings after the renames.

Two coverage gaps the deletions exposed are now closed:

• tests/test_cli_exit.py pins the hard-exit guard, which had no tests at all.
• The TUI chat e2e gains a multi-turn case. It reads the persisted session rather than the screen, because a prompt is echoed into the transcript the moment it is typed -- screen-scraping for a planted word passes even when history is broken.

Three timing bugs fixed in the TUI e2e tests. They keyed off the banner, which paints about seven seconds before the app accepts input, and submitted the moment text was typed. Dropped prompts made the pipeline look dead in one test, and made a negative assertion (assert not leak_detected) pass vacuously in another -- no prompt means no streaming means no leak.

Gaps in the two features that landed during review, small enough to close here: the session.py wiring that fills update_available / update_command (removing it left the suite green), _upgrade_command_works (a fixture stubbed it to True everywhere), an episodeSummary case with an empty test body, and clipToWidth / hasMeaningfulReasoning, which had no coverage at all. The larger items from that audit are tracked in #229.

Upstream review attributions dropped from ui-tui comments. 36 mentions of another project's review thread (#19835 as "Copilot round-N review"), plus #18994, #19194 and #14045 -- none resolve to anything here, and AGENTS.md section 1.1 rules out comments pointing at information only visible elsewhere. The rationale each carried is kept; four describe() titles lose a trailing (#18994). Three went further and documented a python counterpart that does not exist: platform.ts credited key-spelling normalization to raven_cli/voice.py (no such module -- raven/tui_rpc/methods/_stubs.py answers voice.toggle with "voice not supported in Raven v0.1"), and useConfigSync.ts plus its test credited an interrupt default to raven_cli/config.py and tui_gateway/server.py::_load_busy_input_mode (busy_input_mode appears nowhere in python; neither module exists).

One flaky test fixed, caught by this PR's own CI run. test_stdio_no_executor_does_not_raise set command="true" and let connect_mcp_servers really spawn it, so the result depended on how a live child process raced the MCP handshake -- it failed once on CI in the full-suite run while passing 30 out of 30 standalone runs locally. It also asserted only that whatever came back was not a SandboxInitError. The transport now raises instead of running (the idiom its sibling test already uses) and the test asserts the stdio branch was reached.

Review follow-ups, folded in: the hard-exit test now skips rather than asserting no lancedb thread exists process-wide, tests/conftest.py no longer credits the pytest segfault to a lancedb thread the suite never starts (raven imports lancedb nowhere), and the status-bar patterns the two TUI e2e tests wait on live in one module instead of a hand-mirrored copy per file.

Every new or changed assertion was checked by mutating the code under test and confirming the assertion fails.

Type

• Others

Test files, test configuration, and comments. The only non-test files touched are ui-tui comments and four describe() titles -- no behaviour change.

Verification

uv run --all-extras pytest -q
# 4654 passed, 25 deselected

cd ui-tui && npx vitest run
# 936 passed, no skips

uv run --all-extras pytest tests/tui/autotest -m e2e -q
# passes locally with tui-use installed and ui-tui built

bash .claude/scripts/preflight_ci.sh
# ALL PREFLIGHT CHECKS PASSED (including the TUI lane)

Mutation checks, each confirming the new assertion fails when the code under test is broken: the session.py nudge wiring removed; clipToWidth replaced with identity; hasMeaningfulReasoning forced to true; a never-sent prompt added to the multi-turn session assertion.

• Relevant tests pass locally
• Relevant lint / type checks pass locally

Risk

• Security impact considered
• Backward compatibility considered
• Rollback path is clear for risky changes

No behaviour changes: the non-test diff is comments only. The renames touch git history for 19 test files; #208 is the only open PR affected and has been notified.

Two known items are deliberately left alone: test_dogfood_whitelist.py is flaky when the whole file runs (tracked in #228, unchanged here), and the lancedb_finalization_hazard gate in commands.py never fires today -- its test docstring says so explicitly, so the coverage does not read as an argument for keeping it.

Related Issues

Closes #223

[0xKT]

Reviewed the whole diff (merge-base..head, 51 files, renames resolved with -M) in a
separate worktree. Verdict: approve / mergeable. No before-merge findings. Four
items below, all follow-up or nit; R4 from my first pass is withdrawn.

Verified, so no need to re-check

• Renames are complete. Grepped all 20 old basenames across the whole repo
  (docs/, .github/, Makefile, scripts/, benchmarks/): zero hits. The
  detect-private-key exclude in .pre-commit-config.yaml:25 and the three docs
  paths were updated in the same commit.
• Both deletions are justified. test_chat_e2e.py was a module-level skip over
  three NotImplementedError stubs; tui_commands.py:520 does instantiate
  SubscriptionEmitter, and docs/openspec/changes/tui-chat/ no longer exists.
  test_tui_exit_e2e.py asserted lancedb_finalization_hazard() is true, and the
  probe keys on a live thread name that nothing opens today, so the test could only
  fail (rc 43) or skip. tests/test_cli_exit.py is a strictly better replacement.
• The six dropped importorskip guards were all on hard dependencies. httpx
  and everos[multimodal]==1.1.3 are in [project.dependencies];
  raven.plugin.memory.everos is bundled in-repo.
• New API references all exist. SessionManager.find_most_recent_chat_id /
  peek / get_history (raven/session/manager.py:192,402,88), clipToWidth
  (ui-tui/src/lib/text.ts:65), hasMeaningfulReasoning
  (ui-tui/src/lib/reasoning.ts:51).
• _WORKING_RE matches the source of truth. All 15 verbs are identical to
  VERBS in ui-tui/src/content/verbs.ts:22-38.
• The removed locals and dead blocks are genuinely dead (the response block in
  test_task_discoverer.py was shadowed by the reassignment below it).
• Branch name, commit type, header length, Co-authored-by trailer, and ASCII-only
  in both the commit message and this description: all compliant.

---

R1 - the hazard probe test asserts process-global state

• Tier: follow-up
• Where: tests/test_cli_exit.py:28
• Claim: assert not any(t.name == _LANCEDB_THREAD for t in threading.enumerate())
  asserts a property of the whole process, not of the unit under test.
• Why it matters: it holds today (full suite is green), but any future test that
  opens a lancedb connection earlier in collection order flips it red, and the
  failure message will read as "the exit guard broke" rather than "something opened
  lancedb". tests/conftest.py:15 also still documents the opposite ("native
  runtimes pulled in by the suite (lancedb's Rust/tokio thread ...)"), so the two
  docstrings now contradict each other.
• Fix: turn the precondition into a guard rather than an assertion, and reconcile
  the conftest comment with what is actually true today:

def test_hazard_false_without_the_lancedb_thread() -> None:
    if any(t.name == _LANCEDB_THREAD for t in threading.enumerate()):
        pytest.skip("a lancedb connection is already open in this process")
    assert lancedb_finalization_hazard() is False

• Verify: uv run pytest tests/test_cli_exit.py -q standalone and inside
  uv run --all-extras pytest -q; both still pass, and the test skips rather than
  fails when a thread of that name is alive at collection time.

R2 - the coverage pins a gate that never fires

• Tier: follow-up (separate PR - out of scope here, this one is test-only)
• Where: raven/cli/commands.py:177-178
• Claim: if lancedb_finalization_hazard(): flush_and_hard_exit(code) is dead.
  The probe returns False in every configuration today (this PR's own docstring says
  so), and the only live caller of flush_and_hard_exit is tests/conftest.py:32.
• Why it matters: not a defect in this PR - the scoping call to leave production
  code alone is right. Flagging it so the new coverage does not later become the
  argument for keeping the branch: an audit that proves a gate is dormant should end
  in deleting the gate, not in pinning it.
• Fix: open an issue to delete the branch in commands.py and keep
  flush_and_hard_exit for the pytest hook. If it is kept deliberately (e.g. a
  planned embedded-lancedb mode), say so in _exit.py's module docstring, which
  currently describes the gate as load-bearing.
• Verify: n/a for this PR.

R3 - the working-verb regex is now duplicated verbatim

• Tier: nit
• Where: tests/tui/autotest/tests/test_e2e_streaming_no_log_overlay.py:40-48
  and tests/tui/autotest/tests/test_e2e_raven_tui_chat.py:43-51
• Claim: both files now carry the same hand-mirrored _READY_RE / _WORKING_RE
  pair, each with a "keep in sync with verbs.ts" comment.
• Why it matters: two hand-maintained copies of one source of truth; a third
  arrives with the next e2e file, and a verb-pool edit in verbs.ts has to find all
  of them.
• Fix: lift both regexes into one module under tests/tui/autotest/ and import
  them in the two test files.
• Verify: uv run --all-extras pytest tests/tui/autotest -m e2e -q locally
  (tui-use installed, ui-tui/dist built).

R4 - withdrawn

I first read _run_turn's harness.screen() polling
(tests/tui/autotest/tests/test_e2e_raven_tui_chat.py:130-139) as a flake vector
versus harness.wait(). It is not, on two counts: the same loop already exists on
main at test_tui_chat_round_trip (so the helper follows the in-file idiom rather
than introducing one), and the loop has to race two patterns at once -
model_not_available routes to skip, the working verb routes to continue - which a
single-pattern wait() cannot express. No action.

R5 - commit body and description numbers disagree

• Tier: no action
• Claim: the commit body says "Renamed 18 files" / "4615 passed"; this
  description says 19 / 4654. The rename count is 19.
• Why it matters: it does not. The repo squash-merges with
  squash_merge_commit_message=PR_BODY, so the correct numbers are what land on
  main. Noting it only in case the commit is ever amended.

[0xKT]

Re-reviewed the full diff at 9c03caa (not just the delta). R1 and R3 are fixed as
suggested, R4 stays withdrawn, R5 needs nothing. One new finding from the third
commit, cosmetic. Approve-ready either way - it is a comment, so merging with it is
also fine.

R7 - stray character left by a comment rewrap

• Tier: nit
• Where: ui-tui/src/lib/platform.ts:225
• Claim: joining the last two lines of the parseVoiceRecordKey doc comment kept
  the continuation line's leading marker as /:

 * back to the documented Ctrl+B default so a typo never silently
 * disables the shortcut. / */

• Why it matters: harmless to tsc, prettier and vitest (it is comment text, which
  is why CI is green), but it is visible garbage in the one commit dedicated to
  cleaning comments up, and it lands on main until somebody next edits that block.
• Fix: drop the / so the block ends disables the shortcut. */.
• Verify: cd ui-tui && npx tsc --noEmit && npx prettier --check "src/**/*.{ts,tsx}".
  I grepped the same pattern across all 11 files this commit touches; this is the only
  occurrence.

Verified on this round, so no need to re-check

• cd591a6 (R1/R3). test_cli_exit.py:28 now skips instead of asserting
  process-global thread state, and tests/conftest.py no longer credits the pytest
  segfault to a lancedb thread the suite never starts. The status-bar patterns live in
  tests/tui/autotest/statusbar.py and both e2e files import them; re is still used
  in both, so nothing was left dangling.
• d8f3c86 (slash routing). Every claim holds: raven_cli is gone repo-wide (no
  COMMAND_REGISTRY anywhere), createSlashHandler.ts:45-51 is the right anchor,
  findSlashCommand is byName.get(name.toLowerCase()) over names plus aliases, and
  all six NATIVE_MUTATING_COMMANDS names are in the local registry - so the dropped
  native branch could not change an outcome, and "native" was never a route the
  implementation had. The negative and alias fixtures are real: no local channels
  command, bg is an alias of background (session.ts:46-48), model exists
  (session.ts:70).
• 9c03caa really is comments only. Read every hunk of platform.test.ts (45
  lines), useConfigSync.test.ts (31), createSlashHandler.test.ts (19) and
  session.ts (22): no assertion added, changed or removed, and session.ts's
  const parsed = r.record_key ? ... : undefined is untouched.
• The sweep is complete. Zero repo-wide hits for 19835, 18994, 19194 or
  14045, and zero raven_cli / tui_gateway references left in ui-tui/src. The
  remaining Copilot matches are all GitHub Copilot the provider.
• The three deleted python back-references were genuinely fictional. No
  raven_cli or tui_gateway module exists, busy_input_mode appears nowhere in
  python, and raven/tui_rpc/methods/_stubs.py answers voice.toggle with "voice not
  supported in Raven v0.1".
• R2 unchanged and still correct. raven/cli/commands.py:177-178 cannot fire:
  nothing under raven/ imports lancedb (all hits are prose), and everos runs
  out-of-process via subprocess.Popen (_server.py:56), so its lancedb thread never
  lives in the CLI process. Still worth a separate PR, still out of scope here.

Residual risk, stated for the record

CI does not run tests/integration/* (deselected) or tests/tui/autotest (needs
tui-use and a reachable model). So the two deletions, the xfail change in
test_everos_backend_e2e.py, the new multi-turn e2e and the three timing fixes rest
on the local runs reported in the description rather than on a machine. That is the
repo's existing CI gap, not something this PR introduced.