| name | task-inspection |
| description | Diagnose the target spacecraft using multimodal evidence. Use for inspection-diagnosis tasks. |
| metadata | {"skill_kind":"task","routing_summary":"Use when the mission is to identify spacecraft type, visible components, and status from observation evidence.","routing_keywords":["inspect","diagnosis","diagnose","identify components","summarize status","spacecraft type"]} |
Task Goal
Collect enough evidence to describe the target spacecraft's type, visible components, status, and likely function, then call terminate_navigation(...) with a concise diagnosis.
Inspection Strategy
- Begin with the current image and the strongest currently exposed visual evidence to understand the target roughly.
- Use local visual tools to inspect visible structure before changing viewpoint.
- Use
set_attitude() when a different angle is needed to disambiguate components or status.
- Use translation only when the current viewpoint cannot support the diagnosis and a new observation position is necessary.
- Do not assume a specific geometry tool is present; complete the diagnosis with the currently exposed evidence.
- If the
[Sensor] LiDAR block shows data, use it only as a standoff or safety cue, not as evidence for semantic diagnosis.
- Once the evidence is sufficient, stop instead of over-collecting redundant observations.
Output Expectation
Your final terminate_navigation(reason=...) must cover these five aspects:
- Type — what class of spacecraft is this (e.g. CubeSat, large bus, etc.)
- Key Components — list visible structural elements (solar panels and count, antennas, thrusters, thermal blankets, sensors, etc.)
- Visual Appearance — overall shape, color, approximate size, and surface texture
- Function — likely mission or purpose based on observed evidence
- Status — operational condition (nominal, damaged, missing parts, etc.)
Include all five in a single concise paragraph or structured list within the termination reason string.