The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the pyspatialml-pico utility via the uv package manager (uv tool install --upgrade --python 3.13 --system-certs pyspatialml-pico). This is a documented setup requirement for SpatialML model verification and visualization, targeting a well-known service registry.
[COMMAND_EXECUTION]: The skill facilitates the execution of various CLI commands, including shell passthrough (pico-cli shell) and ADB operations (pico-cli adb). It defines safety conventions, such as preferring high-level command wrappers over raw shell access and requiring explicit user confirmation before executing destructive commands like file removal or emulator deletion.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from external CLI tools and devices, which represents an indirect prompt injection surface. * Ingestion points: Command output (JSON and plain text) from pico-cli and adb subcommands. * Boundary markers: Instructions strongly prefer the use of --format json for all diagnostic and inspection commands to ensure structured data handling. * Capability inventory: File system operations (push/pull/mkdir/rm), shell execution, and application lifecycle management. * Sanitization: Relies on the agent's interpretation of structured tool output and explicit device targeting to mitigate risks.

pico-cli