@@ -8,16 +8,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
88## [ Unreleased]
99
1010### Added
11- - ** Slop Detection Pipeline Architecture** - 3-phase detection pipeline with certainty-tagged findings (#107 )
11+ - ** Slop Detection Pipeline Architecture** - 3-phase detection pipeline with certainty-tagged findings (#107 , # 116 )
1212 - ** Phase 1** (always runs): Built-in regex patterns + multi-pass analyzers
1313 - ** Phase 2** (optional): CLI tool integration (jscpd, madge, escomplex) - if available
14- - ** Phase 3** : LLM handoff with structured findings
14+ - ** Phase 3** : LLM handoff with structured findings via ` formatHandoffPrompt() `
1515 - Certainty levels: HIGH (regex), MEDIUM (multi-pass), LOW (CLI tools)
1616 - Thoroughness levels: quick (regex only), normal (+multi-pass), deep (+CLI)
1717 - Mode inheritance from deslop-around: report (analyze only) vs apply (fix issues)
1818 - New ` runPipeline() ` function in lib/patterns/pipeline.js
19+ - New ` formatHandoffPrompt() ` for token-efficient LLM handoff (compact output grouped by certainty)
1920 - New lib/patterns/cli-enhancers.js for optional tool detection
20- - deslop-work.md agent updated to use pipeline orchestrator
2121 - Graceful degradation when CLI tools not installed
2222
2323- ** Buzzword Inflation Detection** - New project-level analyzer for ` /deslop-around ` command (#113 )
@@ -109,6 +109,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
109109 - ` speculative_generality_empty_interface ` : Empty TypeScript interfaces
110110 - All patterns have ReDoS-safe regex and comprehensive test coverage
111111
112+ ### Changed
113+ - ** deslop-work Agent Refactor** - Rewrote from pseudo-JavaScript to explicit Bash/Read/Grep tool usage (#116 )
114+ - Now uses pipeline orchestrator (` runPipeline() ` ) instead of inline pattern matching
115+ - Certainty-based decision making: HIGH (auto-fix), MEDIUM (verify context), LOW (investigate)
116+ - Structured handoff via ` formatHandoffPrompt() ` reduces agent prompt verbosity
117+ - Clearer separation: pipeline collects findings, agent makes decisions
118+ - Improved maintainability with declarative tool instructions
119+ - ** deslop-around Command** - Enhanced documentation for mode usage and pattern library (#116 )
120+ - Clarified report vs apply mode behavior
121+ - Added references to pattern library categories
122+ - Updated code smell detection section with latest patterns
123+
112124### Fixed
113125- ** findMatchingBrace** - Now skips comments to avoid breaking on quotes/apostrophes in comment text (e.g., "it's", "we're")
114126- ** Reality Check Output Size** - Condensed collector output to ~ 700 lines/~ 4.5k tokens (was thousands of lines)
0 commit comments