The planner uses SHOP algorithm (the difference is that I don't support operators' add atoms and delete atoms lists)
Domains descriptions use Lisp's S-expressions similar to SHOP and JSHOP. Planning domains are translated to Python code using planner compilation technique as in [JSHOP] (http://www.cs.umd.edu/projects/shop/).
Specific domain description language (methods and branches) is modelled after and influenced by Guerrilla Games' presentation on their use of HTN.
The project is an experiment created for AI Sandbox's Capture The Flag Competition.
The bot I've submitted for the competition took the 1st place in the final phase: http://aisandbox.com/announcement/phase3-results/
You can find some very basic example domains under planner/examples directory.
The code is in public domain.