Find untested code with a coverage audit, then generate pytest tests to close the gaps

Failure modes

Symptom	Cause	Fix	Severity
`RuntimeWarning: coroutine 'execute' was never awaited`	`execute` called without `await`	It is a coroutine — `await` it or use `asyncio.run`	high
`WorkflowResult.success` is `False`, `error` is `"path argument is required"`	`execute` called with empty or missing `path`	Pass a non-empty `path`	high
`DeprecationWarning: ...execute(src_path=...) is deprecated`	The audit was called with the legacy `src_path` kwarg	Use `path=` instead (the audit still runs)	low
`error` reads `"Agent SDK unavailable: ..."`	`claude_agent_sdk` is not importable	Install the Agent SDK dependency for the environment	high
`error` reads `"Agent SDK connection failed: ..."`	A `ConnectionError` / `TimeoutError` reaching the SDK	Check connectivity / retry	medium
Generated tests don't pass as-is	Generation is a predictive starting point	Review, adjust, and run them before committing	medium
Audit finding looks like a false positive	Findings are LLM predictions, not verified defects	Confirm against the cited file/line before acting	medium

The async call is easy to get wrong. execute is a coroutine on every smart-test workflow. Forgetting to await it is the single most common mistake.
Two slugs, one feature. The skill / topic is smart-test, but the CLI slugs are test-audit and test-gen. And a same-named repo-level skill (.claude/skills/smart-test) does something different — it runs your diff's affected tests.
Generation is a draft. test-gen writes a starting point. Run and review the output; a generated test that imports the wrong symbol or asserts the wrong value is on you to catch.

Confirm you are awaiting: result = await TestAuditWorkflow().execute(path="src/") inside an async def or asyncio.run.
Check result.success; if False, read result.error and result.error_type.
For a CLI "unknown workflow" error, confirm the slug is test-audit or test-gen (not smart-test).
On an SDK error, inspect result.metadata for the captured SDK error fields.
Confirm the scope: result.metadata echoes the run's path / src_path, depth, and max_turns.