🪣 Slop-Mop
Slop-mop is a longitudinal force multiplier for humans building with AI. It
optimizes repositories toward long-term maintainability and overall throughput
in three ways:
- enforcing "no-brainer" SOPs (linting, typing, structure, test coverage, etc.)
- catching bullshit (bogus tests, disabled safety checks, ignored feedback, etc.)
- providing "greased rails" for agents to ride along — protocol-like instruction
sequences for both humans and AI that make PR review, comment triage, and
remediation fast and thorough
Coding agents optimize for apparent completion — the nearest green checkmark,
not the right one. They act like balls rolling downhill, and left to their own
momentum they settle in the shallowest local minimum: untested claims, coverage
gamed by a true is true, a silenced gate, a git commit --no-verify. The code
runs. The PR looks clean. The slop is already in — junk DNA that threatens the
code's offspring down the line.
Slop-mop keeps the rule outside the loop. Your standards live in external
gates, not inside the agent's reward function, so they hold even when reward
pressure is high — and following the rail becomes the shortest path to the
reward instead of a wall to climb. Refit carves the initial terrain,
swab/scour/buff keep the gradient pointed at maintainable code, and
wake-angry-drunk-captain blocks fake progress when only a human can break the
tie.
The verbs are deliberately nautical — swab, scour, buff, sail,
barnacle. Novel tokens from naval practice don't come with a million training
examples of how to weasel around them, which helps keep models out of dangerous
eddies.
This is harm reduction, not prevention. A determined model will still find a
seam when the reward pressure is high enough; the honest claim is narrower —
more catches than misses, over time. That is enough to keep a codebase
navigable. It is purposefully opinionated, because structure begets
adherence to best practices.
Quick Start
Install it and set up a repo (omit [all] for the framework only — gates whose
tools like black, pyright, or pytest are missing will say so):
pipx install slopmop[all]
sm init
Inherited an existing codebase? Run refit first — it builds a remediation plan
and walks you gate-by-gate until the repo is clean enough for the daily loop:
sm refit --start
sm refit --iterate
sm refit --finish
Then it's just the loop: sm swab while you work, sm scour before a PR,
sm buff after CI or review feedback lands. Not sure what's next? sm sail
reads the workflow state and runs the right verb. See The Loop for
the full table, or Baselines if you need to defer the cleanup.
Use with Claude
Slop-mop ships as a Claude plugin: a skill that auto-triggers on remediation
prompts, plus eight slash commands (/sm-init, /sm-refit, /sm-sail,
/sm-swab, /sm-scour, /sm-buff, /sm-barnacle,
/sm-wake-angry-drunk-captain). Install once and sm is available in every
repo — no per-repo sm agent install required.
In Claude Code or Cowork:
/plugin marketplace add ScienceIsNeato/slop-mop
/plugin install slopmop
