| Performer and Work are named | A course, tool, or vague mastery goal drives the project. | State the holder, receiving Work, and decision first. |
| Capability is attributed to the right holder | Person, device, robot, pair, ensemble, and combined whole collapse. | Keep their capability claims separate and test the holder used by the decision. |
| Target and protected conditions precede intervention | Available exercises become the objective. | Derive the smallest capability demand from representative Work. |
| Contribution is specific but not isolated falsely | A local gain damages performance elsewhere. | Keep simultaneous interactions and protected trade-offs visible. |
| Development Method fits the holder | Human training language is copied into robot, AI, equipped, or collective cases. | Use the direct holder-specific Method and evidence. |
| Practice prepares later contexts | Repetition in one context masquerades as transfer. | Vary relevant contexts, cues, transition history, and delayed use where the claim requires it. |
| Context inference remains a hypothesis | Apparent forgetting receives one automatic explanation. | Use return-to-context and discriminating probes; retain rivals. |
| Representative Work is performed | Activity, credentials, or exercise success replace capability. | Test the target and protected conditions in receiving-like Work. |
| Causation is not invented | Improvement is credited automatically to teacher, exercise, model, or device. | Keep Work, before/after evidence, capability, and causal claim separate. |
| Result has limits and a reopen rule | One success becomes universal mastery. | State envelope, unsupported conditions, and currentness trigger. |