What problem does it solve? Codebases accumulate duplication, dead code, weak typing, defensive clutter, and AI-generation tells that degrade maintainability, and manually auditing a large codebase or PR for all of these at once is slow and inconsistent. ## Core Features & Use Cases - Seven parallel analysis agents: Deduplication, type consolidation, unused code, weak types, defensive programming, legacy code, and AI slop / over-nesting / style drift, all reporting findings in a shared JSON format. - Scoped sweeps: Run against the full codebase, a directory, a PR number (resolved via gh), or a branch diff, with automatic excludes for build output, vendored code, and lockfiles. - Safe apply workflow: Findings are consolidated into deslop-report.md, reviewed with the user, then applied by category in isolated git worktrees with build, typecheck, and tests run between merges. - Use Case: Before merging a large feature branch, run a deslop sweep on the diff to strip AI-generated comments, flatten over-nested conditionals, and remove unused exports, then apply the approved fixes on a dedicated cleanup branch. ## Quick Start Ask the assistant to deslop the current branch or a directory such as src/api and review the consolidated report before applying fixes.