mopthing
Planned · the commands are the design
The slop, measured.
An agent stops when the tests pass, not when the code is good. mopthing runs the analysers and hands the agent what they find.
- 1 command, every repo
- 1 format for every analyser
- 30 CRAP, the line
- 0 old debt that blocks
Features
Findings the agent actually sees.
-
Every analyser, one format
Complexity, coverage and duplication from each language's own tools, read into one table.
-
One score
CRAP squares complexity and cubes the share left uncovered, so tangled untested code stands out.
-
A ratchet
A baseline per repo means old debt is recorded, not failed. Only new or worse code goes back.
How it works
Into the agent's loop.
-
1
Run
The analysers stay upstream and mopthing reads what they print, so nothing is reimplemented.
-
2
Check
After the agent's turn, only changed code is checked, so it fails where the agent can act.
-
3
Hand back
A failing check exits non-zero with what to do, so its output becomes the agent's next prompt.
Who built it
Built by Rytter.
I'm Jakob Rytter. I build software and run it, from the server to the screen, and mopthing is how my agents' code will get checked.