Code Journey Rework: From Broken Scoring to Real Bug-Finding
6 sections
Overview
Transform the code-journey system from a broken scoring/analysis pipeline that produces false confidence (18/20 journeys score 10.0 while real memory leaks go undetected) into a grounded bug-finding system that composes the project's proven diagnostic tools (ORI_CHECK_LEAKS=1, rc-stats.sh, codegen-audit.sh, dual-exec-verify.sh) with AI-driven IR analysis that has historically found real bugs (double-free on [str] elements, closure Idx leak, aggregate materialization 9:1 fix, landing pad over-generation). Results are captured as canonical JSON that feeds both human-readable markdown and website visualization. When bugs are found, they are automatically filed via /add-bug.
Planned
6 sections
Tear Down Broken Pipeline
0/28 tasks
JSON Results Schema
0/18 tasks
Orchestrator Skill
0/34 tasks
Diagnostic Tool JSON Output
0/30 tasks
Journey Corpus Expansion
0/28 tasks
Bug Integration & Markdown Generation
0/26 tasks