exploration-guided-unfamiliar-app-adaptation

# Skill: exploration-guided-unfamiliar-app-adaptation ## 1. Capability Definition & Real Case * **Professional Definition**: The capability to acquire actionable knowledge about an unfamiliar GUI environment through prior exploration and then use that knowledge to solve downstream tasks in the same application. The agent must convert exploration evidence into app-specific navigation priors, functional understanding, and route selection rather than over-relying on generic cross-app assumptions. * **Dimension Hierarchy**: GUI Perception and Environment Modeling->Environment Understanding->exploration-guided-unfamiliar-app-adaptation ### Real Case **[Case 1]** * **Initial Environment**: A messaging-and-payment super-app is open on its home screen. The agent has access to an exploration trace showing how wallet, payment, and subscription-management pages are organized, but the target feature is not exposed from the initial view and the app’s information architecture does not mirror common Western payment apps. * **Real Question**: Please cancel auto-payment service in WeChat. * **Real Trajectory**: Consult the exploration trace, infer that the relevant path runs through wallet and payment management rather than a generic settings page, navigate to the auto-renewal area, identify the active service, and cancel it. * **Real Answer**: The auto-payment service is canceled. * **Why this demonstrates the capability**: The task is difficult precisely because prior experience with other payment apps can suggest the wrong route. Success depends on mining the exploration evidence for app-specific structure, then translating that evidence into a correct action sequence in the unfamiliar app. That is the defining pattern of exploration-guided adaptation. ## Pipeline Execution Instructions To synthesize data for this capability, you must strictly follow a 3-phase pipeline. **Do not hallucinate steps.** Read the corresponding reference file for each phase sequentially: 1. **Phase 1: Environment Exploration** Read the exploration guidelines to discover raw knowledge seeds: `references/EXPLORATION.md` 2. **Phase 2: Trajectory Selection** Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: `references/SELECTION.md` 3. **Phase 3: Data Synthesis** Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: `references/SYNTHESIS.md`