3-phase AI slop detection: regex patterns (HIGH), multi-pass analyzers (MEDIUM), CLI tools (LOW)
npx claudepluginhub agent-sh/deslopAI slop cleanup with minimal diffs and behavior preservation
Claude Code marketplace entries for the plugin-safe Antigravity Awesome Skills library and its compatible editorial bundles.
Production-ready workflow orchestration with 79 focused plugins, 184 specialized agents, and 150 skills - optimized for granular installation and minimal token usage
Browser automation for AI agents
Detect and remove AI-generated slop from codebases with certainty-based findings and safe auto-fixes.
AI coding tools leave behind debug statements, placeholder text, empty catch blocks, over-commented code, and dead abstractions. Manual cleanup is tedious and error-prone. deslop runs a 3-phase detection pipeline that categorizes every finding by certainty level - HIGH findings get auto-fixed, MEDIUM findings get flagged for review, and LOW findings are reported without action. Behavior is preserved. Diffs are minimal.
Use cases:
agentsys install deslop
# Report slop findings (no changes made)
/deslop
# Auto-fix HIGH certainty findings
/deslop apply
# Scan only changed files in current branch
/deslop report --scope=diff
# Fix up to 10 findings in a specific directory
/deslop apply src/ 10
deslop uses a 3-phase detection pipeline with increasing analysis depth:
Phase 1 - Regex patterns (HIGH certainty). Fast pattern matching for console.log, print(), dbg!(), TODO/FIXME markers, empty catch blocks, hardcoded secrets, trailing whitespace, and mixed indentation. These are safe to auto-fix.
Phase 2 - Multi-pass analyzers (MEDIUM certainty). Structural analysis for doc-to-code ratio problems, verbose over-commenting, over-engineering, buzzword inflation, dead code after return/throw, and stub functions. These need human review.
Phase 3 - CLI tools (LOW certainty, optional). Runs external tools when available - jscpd for duplication, madge for circular dependencies, eslint/pylint/clippy/golangci-lint for language-specific issues. Findings are flagged but not auto-fixed.
Thoroughness levels control which phases run:
| Level | Phases | Speed |
|---|---|---|
quick | Phase 1 only | Seconds |
normal (default) | Phase 1 + 2 | Seconds |
deep | Phase 1 + 2 + 3 | Depends on CLI tools |
Repo-intel integration - when repo-intel data is available, deslop targets AI-written files first (recent-ai query) and escalates MEDIUM findings to HIGH for files with no test coverage (test-gaps query).
| Level | Meaning | Action |
|---|---|---|
| HIGH | Definitely slop - safe to remove | Auto-fixed in apply mode |
| MEDIUM | Likely slop - needs context | Flagged for review |
| LOW | Possible slop - context-dependent | Reported only |
/deslop
/deslop report --scope=diff
/deslop report --thoroughness=deep
Outputs a prioritized table of findings with certainty levels and suggested fixes. No files are modified.
/deslop apply
/deslop apply --scope=diff
/deslop apply src/ 10
Auto-fixes all HIGH certainty findings, then runs the project's test suite. If tests fail, all changes are rolled back with git restore . and the failing fix is reported.
all (default) - scan entire codebasediff - only files changed in current branch<path> - specific directory or file/deslop --thoroughness=quick # Phase 1 only
/deslop --thoroughness=normal # Phase 1 + 2 (default)
/deslop --thoroughness=deep # All phases
JavaScript/TypeScript, Python, Rust, Go, Java/Kotlin, C/C++.
MIT