cyberneticlibrary

Classify memory safety issues

phase-h-classify-issuesworkflowsetup L20
Microck/bun-rust-port-claude-artifacts
What it does

Classify GitHub memory safety issues, synthesize reports

Best for

Prioritizing crash/memory-safety issues in a Zig→Rust migration by classifying GitHub issues at scale with Rust-port impact assessment.

Inputs
  • · Issue index JSON (number, title, path)
  • · Repository path
  • · Batch size (default 20)
Outputs
  • · Classification results (category, confidence, area, version, repro, likely_fixed_by_port)
  • · CRASH_LEAK_ISSUES.md (markdown report)
  • · crash-leak-issues.csv (CSV for downstream analysis)
Requires
  • · GitHub API (read-only, no mutations)
  • · Claude agents (per-batch classification + synthesis)
Preconditions
  • · Index JSON at ${REPO}/tmp/crash-leak-issues/index.json
  • · Issue JSON files readable at specified paths
  • · Multi-phase: Classify → Synth
Failure modes
  • · Missing index → throws error
  • · Confidence varies (high/medium/low); low-confidence results unreliable
  • · Classification mislabels logic bugs as memory bugs
  • · likely_fixed_by_port guesses about Zig rewrite impact; requires manual review
Trust signals
  • · 10-category enum (crash, leak, oom, hang, uaf, double-free, assert, not-memory, duplicate, needs-repro)
  • · Confidence levels (high/medium/low)
  • · Batched per ~20 issues (context-aware)
  • · Markdown + CSV export
  • · Area classification (fetch/http/bundler/install/fs/...)