lock-e2e

Freeze a passing action into a locked e2e regression test. A locked test is a copy of the action in .rn-agent/e2e/<id>.yaml that replays strict — self-repair is disabled, so any drift fails loudly instead of being absorbed. Locked tests are what cdp_run_e2e_suite (and the observe UI’s E2E tab) run as the project’s regression suite.

Usage

/rn-dev-agent:lock-e2e <action-name> [--relock]

How it works

1. The action is replayed once in strict mode (no repair) via cdp_lock_e2e_test.
2. Only if that run passes is the flow frozen to .rn-agent/e2e/<id>.yaml.
3. From then on it is included in cdp_run_e2e_suite runs.

--relock re-freezes an already-locked test after the underlying action changed (for example after a re-record).

Failure modes

Result	Meaning
STRICT_RUN_FAILED	The action doesn’t currently pass without repair. Run /run-action first so self-repair can patch drift, then lock again.
MISSING_PARAMS	The action has ${KEY} placeholders that aren’t covered. Provide values in .rn-agent/e2e.config.json — per test under tests.<actionId>.params, or shared under defaults.params — and retry.

Why strict?

Actions absorb small UI drift by design — useful for prologues, wrong for regression tests. Locking captures a moment where the flow passed with zero assistance, so a later failure is a real signal, not repaired-over noise.

See also

• Actions — record and replay the flows you’ll want to lock
• /run-action — repair-enabled replay for unlocking STRICT_RUN_FAILED
• /observe — run the locked suite from the browser’s E2E tab