You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Add bhom-python Claude Code skill for writing, testing and reviewing Python #264
Add a Claude Code skill to Python_Toolkit so AI-assisted development writes, tests and reviews Python to the BHoM standard.
The skill lives in .claude/skills/bhom-python/ and builds on the published guidance in Python_Engine/Python/README.md. It contains:
SKILL.md - overview, golden rules and decision tree
SKILL-IMPLEMENTATION.md - package layout, packaging and style rules
SKILL-FRAMEWORK.md - bhom_wrapper.bhom_callable + run_wrapped as the route from C# into Python, BHoMObject/IObject and BHoM JSON, logging, bhom_analytics and pythonnet
SKILL-TESTING.md - pytest standards, including tests for wrapped (C#-callable) methods
SKILL-REVIEW.md - compliance rule catalogue and review report format
Expected behaviour: when working in this repo, Claude Code loads the skill for Python work, and its code and reviews follow the rules above.
Open decisions to settle during review:
Copyright header on .py files
Test location (Python_Engine/Python/tests/ vs .ci/unit_tests/python/)
Identifier convention for wrapped methods
A black/pylint configuration and a Python CI step
While drafting the skill, these framework issues came up. They can be split into separate issues if preferred:
bhom_callable fails when the input JSON root is a BHoM-serialised CustomObject or Dictionary, because BHoM_Guid / _t / _bhomVersion are passed as keyword arguments
bhom_callable uses a bare except:, and invalid JSON then raises UnboundLocalError
bhom/wrapped/_test_wrapped.py registers a "test" method that prints to stdout
CONSOLE_LOGGER writes to stdout, the same stream as wrapped results
Activity