Make an existing tmula scenario actually runnable and safe — substitute literal path params, wire auth headers from the credential pool, filter out destructive operations, and add dependencies / weights / finding thresholds. Use when you already have a scenario.json (from tmula-scaffold, hand-written, or pasted) that won't run cleanly or safely yet. Does NOT send traffic — that's tmula-run.
Run a tmula scenario as a load test, safely. Enforces a non-production safety gate, smoke-tests with one user, then executes the load (closed or open arrival model) and reads the findings honestly. Use when you have a runnable json/scenario.json and want to actually send traffic / stress-test / load-test the target. This is the step that sends real HTTP requests.
Scaffold a starting tmula scenario from an API description — an OpenAPI/Swagger doc (URL or file), a HAR capture, or an access log. Use when you have a spec or traffic log and need a tmula scenario.json to start from. Produces a RAW scaffold (linear flow, literal path params, no auth, destructive ops still present); pair with tmula-enrich to make it runnable and safe. Does NOT send traffic — that's tmula-run.
Triage a finished tmula run — reproduce a finding in isolation to tell a functional bug from a load-dependent one, gate the run against a baseline, and wire it into CI. Use after a tmula run produced findings and you want to know which are real, whether anything regressed, or how to make load testing a CI gate. Reproduce sends a small amount of real traffic to replay the failing session.
End-to-end tmula onboarding — take an API all the way from a source (OpenAPI/Swagger URL or file, HAR, or access log) to a safe load test and triaged findings. Orchestrates scaffold → enrich → run → triage with confirmation gates, picking up wherever you already are. Use when you want the whole thing done from a single input ("load-test this API from its swagger, end to end", "tmula 한 방에 돌려줘"). Sends real traffic at the run stage, behind a safety gate.