The anatomy of a 14-hour harness session — what a fully instrumented Claude Code orchestration run actually did, turn by turn, with a receipt for every move
mpm has it’s own agent repository. The man reason is that it enforces a specific workflow that works better for me. Also, agents inherit behaviors, so I have a build process when deploying agents (base agent —> base engineer —> rust engineer etc), which provides more consistent behavior. All in bobmatnyc/claude-mpm - do pull skills from various sources.
Wow, really cool session. Are you writing your own agents or getting them open-sourced?
mpm has it’s own agent repository. The man reason is that it enforces a specific workflow that works better for me. Also, agents inherit behaviors, so I have a build process when deploying agents (base agent —> base engineer —> rust engineer etc), which provides more consistent behavior. All in bobmatnyc/claude-mpm - do pull skills from various sources.
What did this actually produce though?