diff --git a/skills/hcq-hil/SKILL.md b/skills/hcq-hil/SKILL.md index 939bf66..9b18348 100644 --- a/skills/hcq-hil/SKILL.md +++ b/skills/hcq-hil/SKILL.md @@ -103,13 +103,12 @@ controller exposes independent UART-fault evidence. simulator input entities in [the REST reference](references/esphome-rest.md) to vary room temperature, room setpoint, `TSet`, boiler telemetry or ODU water input. Read the allowed range, set one controlled value, verify the - controller outcome and restore the prior value. Do not conflate simulated + controller outcome. Do not conflate simulated boiler telemetry with a physical PT1000 supply-sensor test. - Treat the testcontroller's input-source selection as part of the test state. Read the controller source selectors and use its verified `api_input_*` HTTP endpoints for controller-side room, outdoor, dew-point, demand and enable - scenarios. Verify source selection and freshness after every write, then - restore the previous source and input state. The current API-input surface + scenarios. Verify source selection and freshness after every write. The current API-input surface has no water-supply-temperature input; do not substitute another sensor or claim a supply-temperature API test. - Check the actual behavior and relevant counters; do not treat a successful @@ -123,6 +122,13 @@ controller exposes independent UART-fault evidence. the final simulator state with the result instead. - Still leave an active fault injection disabled: use the fault helper or turn its switch off before ending the current run. +- Do not restore ordinary testcontroller settings, input-source selections or + API-input values after a HIL run unless the user or scenario requests it. + Each new test uploads its own firmware and establishes its own test state. + Firmware uploads can preserve settings, so explicitly set and verify the + inputs and selectors needed by each scenario rather than assuming defaults. + Record relevant final controller settings with the result; restore a value + within a run only when a subsequent assertion needs that baseline. - For a normal “test deze PR”-request, run the read-only smoke test directly. Ask one concise question only if the relevant test needs OTA/flash, reboot, fault injection, a change outside the requested diff, or an action that diff --git a/skills/hcq-hil/references/esphome-rest.md b/skills/hcq-hil/references/esphome-rest.md index b37b1ea..b0aa3a7 100644 --- a/skills/hcq-hil/references/esphome-rest.md +++ b/skills/hcq-hil/references/esphome-rest.md @@ -102,7 +102,9 @@ curl -fsS 'http://openquatt-test.local/select/Room%20Setpoint%20Source?detail=al Use `POST -H 'Content-Length: 0'` and follow every write with a GET. Select `API input` on the matching controller source before interpreting the control -result, then restore the prior source and input state after the test. Example: +result. Leave ordinary source selections and input values in place after the +test; record them and explicitly establish the next scenario's required state. +Do not assume that a firmware upload clears persisted settings. Example: ```sh curl -fsS -X POST -H 'Content-Length: 0' \