slopmop-swab · diff
git:20260311.822c565 to git:20260311.923e572
16 added, 16 removed. Audit A to A.
---
- description: Slop-mop quality-gate skill
+ description: Slop-mop development protocol — gradient descent for repo health
alwaysApply: true
---
- ## Slop-mop skill
+ ## Slop-mop — your development protocol
- Slop-mop (`sm`) is a quality-gate runner for this repository. It provides three verbs that form a continuous validation loop:
+ Slop-mop (`sm`) is a gradient descent tool for this repository. It treats common LLM weaknesses — sloppy formatting, dead code, stale docs, untested changes — as adversarial patterns that cause accidental repo rot. Rather than leaving you to figure out what to validate and when, sm provides greased rails: a fast, repeatable development loop that tells you exactly what to do next.
- | Verb | Purpose | When to use |
- |------|---------|-------------|
- | `sm swab` | Fast validation — runs configured quality gates and auto-fixes | During implementation, after meaningful code changes |
- | `sm scour` | Full PR-readiness check — runs all gates without auto-fix | Before opening or updating a pull request |
- | `sm buff` | Post-PR triage — digests CI results and review feedback | After CI completes or review feedback lands on a PR |
+ **This is your procedural cornerstone.** Follow the swab/scour/buff loop and focus on writing great code.
- ### Required workflow
- 1. Run `sm swab` early and often during implementation.
- 2. If swab reports failures, summarize the failing gates, apply fixes, and re-run until clean.
- 3. Before opening or updating a PR, run `sm scour`.
- 4. After CI completes or review feedback lands, run `sm buff <PR_NUMBER>` and follow the guidance.
+ | Verb | What it does | When to run it |
+ |------|-------------|----------------|
+ | `sm swab` | Fast iterative feedback — catches drift and auto-fixes what it can | After every meaningful code change (run early, run often) |
+ | `sm scour` | Comprehensive pre-PR sweep — surfaces everything before it compounds | Before opening or updating a pull request |
+ | `sm buff` | Post-PR triage — digests CI results and review feedback into actionable steps | After CI completes or review feedback lands |
+ ### The loop
+ 1. **Write code → `sm swab` → fix → repeat.** This is gradient descent — each pass reduces entropy. Keep running until clean.
+ 2. **Before PR → `sm scour`.** One comprehensive pass to catch anything swab missed.
+ 3. **After CI/review → `sm buff <PR_NUMBER>`.** Converts raw feedback into a concrete remediation plan.
+
### Tooling preference
- If MCP tools `sm_swab`, `sm_scour`, `sm_buff` are available (e.g. via a configured MCP server), prefer calling those tools.
- Otherwise, run the CLI commands directly from the project root.
### Safety
- - Never bypass, disable, or silence failing gates as a shortcut.
- - Only run `sm` subcommands needed for the swab/scour/buff loop.
- - If output is large, summarize and cite the most actionable items.
+ - Never bypass, disable, or silence a failing check — that's how repo rot starts.
+ - If output is large, summarize and cite the most actionable items first.