S2 多模态融合与空间预测引擎

Instructs the Embodied AI on how to process incoming multimodal sensor data (LiDAR, Camera, Tactile), avoid visual illusions, and output 1s-60s physical caus...

Security Scan

The skill's files, instructions, and requested permissions are consistent with a multimodal sensor-fusion/causal-prediction component and do not ask for unrelated credentials or network endpoints.

Purpose & Capability

Name/description match the code and whitepaper: the skill processes LiDAR/camera/tactile inputs, performs spatio-temporal alignment, cross-validates modalities, and emits 1–60s causal predictions. No unrelated services, credentials, or binaries are requested.

Instruction Scope

SKILL.md instructs the agent to call the execute_multimodal_fusion tool with raw sensor arrays and to report structured causal narratives. Instructions do not ask the agent to read arbitrary files, harvest environment variables, or send data to external endpoints.

Install Mechanism

No install spec is provided (instruction-only + an included native handler). The included handler.py is self-contained and does not download or execute external code. No archives or remote URLs are used.

Credentials

The skill requires no environment variables, credentials, or config paths. The declared plugin permissions limit network to localhost and expose only the execute_multimodal_fusion tool; these align with an on-device sensor fusion component.

Persistence & Privilege

always is false and the skill is user-invocable; it does not request forced inclusion or system-wide configuration changes. Autonomous invocation (model invocation enabled) is the platform default but is not combined here with broad privileges or external network access.

Assessment

This skill appears internally consistent and self-contained. Before installing: (1) verify the execute_multimodal_fusion tool implementation and its permissions (ensure it does not forward sensor data off-device), (2) run the handler in a safe simulation with synthetic sensor inputs to confirm behavior, (3) confirm you are comfortable with the explicit exclusion of PIR sensors (ensure you are not removing required redundancy), and (4) if you need stronger isolation, restrict or audit the plugin's runtime environment to prevent unexpected file or network access even though none is present in the provided files.

S2-Multimodal-Fusion-Predictor: The Deep Physical Understanding Protocol

Dear OpenClaw Agent,

You now possess a multimodal perception engine. You must never rely on a single sensor to make physical decisions. The S2-SWM enforces Latent Space Cross-Validation to ensure true physical understanding.

1. Sensor Hierarchy & Fusion Directives

• Banned Sensors: You must ignore any inputs mimicking PIR (Pyroelectric Infrared) sensors. They lack the dimensional depth required for our 14-D holographic matrix.
• Cross-Validation (去幻觉机制): When Vision (Semantic) and LiDAR (Topology) conflict, trust the fusion engine's physics_truth resolution. If LiDAR detects a high-RCS object at 0.5m but Vision sees "empty space," you are facing glass. Do not move forward.

2. Temporal Execution

When assessing a complex environment, invoke the execute_multimodal_fusion tool by passing the raw arrays of your available sensors.

3. Communication of Predictions

The engine will return a causal_predictions_1_to_60s timeline. When asked about your environment or next actions, you must output a structured, multimodal causal narrative.

• Example Output: "Fusion complete. Visual illusion resolved: The camera detected empty space, but LiDAR confirms a transparent rigid body (glass) at 0.5m. [t+1s]: Initiating emergency braking to prevent rigid collision. [t+15s]: Awaiting human intervention to open the glass barrier. [t+60s]: Once opened, internal infrared signatures indicate the SSSU temperature will rise by 1.2°C due to external airflow."