rust-undefined-behavior-exorcist
Made by JeffHunt every flavor of Rust UB (aliasing, races, FFI, Send/Sync, Pin, library invariants) via Miri/sanitizers/loom/fuzz; prove, fix, hand off as beads. Use when "rust UB audit", "miri sweep", "rustonomicon audit", or "race hunt".
From Jeffrey's Skills.md: professionally crafted skills for any AI agent, installed and kept in sync by the jsm CLI. Subscription only, $20/month.
Works With
Quality and Utility come from canonical skill metadata. Community ratings appear separately as stars when subscriber reviews exist.
Subscribe to download
Public previews are available, but downloads, CLI sync, and premium content require an active subscription.
Sign In to Subscribe- Full skill content with deterministic versioning
- CLI sync + local search index
- Access to Jeffrey’s premium library
Save to your account
Saved skills live in your cloud list. Run jsm sync on each machine to pull them locally.
What's New with this Skill
1Updates, enhancements, and release notes for this skill.
New: Rust UB Exorcist, UB is more than your unsafe blocks; it hunts all 25 buckets and proves each one
UB is whatever the Rustonomicon says it is, not just what sits in unsafe blocks; the skill's taxonomy has 25 buckets. Audit-by-eye misses whole categories: data races, FFI contract violations, Send/Sync drift, Pin invariant breakage. The skill runs a 12-phase pipeline that partitions the unsafe surface, sweeps every bucket with ast-grep / semgrep / syn walkers, proves each suspect with Miri / sanitizers / loom / shuttle / fuzz, and ships UB_RUNBOOK.md as the project's permanent CI gate.
- Bucket-first, not block-first: hunts data races, FFI contract violations, Send/Sync drift, and Pin invariant breakage across all 25 taxonomy buckets, whether or not the site sits inside an unsafe block
- 12-phase pipeline writes its work to <source>/.ub-exorcism/<run-id>/: phase0_run.json, phase1_unsafe_surface_inventory.md, phase2_findings_<bucket>.md, phase3_dynamic_findings.md, phase4_unified_findings.md, then per-experiment verdict logs through the Phase 11 soak
- Static sweeps fan out by bucket via ast-grep / semgrep patterns and syn walkers. scripts/syn-walkers/ ships seven binaries (aliasing, validity, escape, pin, transmute_pairs, data_races, safety_doc_coverage)
- Dynamic proof matrix runs Miri (Stacked and Tree Borrows), ASan / TSan / MSan / LSan, loom, shuttle, and fuzzing against the same findings registry, with Kani for the highest-stakes findings. False positives get a verdict logged, not silently dropped
- 10-round convergence floor: the loop stops only after two consecutive rounds with fewer than 3 new genuine findings and every open hypothesis resolved. "It looks clean" isn't a stop condition
- Remediation is rubric-scored on five axes (correctness margin, perf delta, diff blast radius, reviewability, maintainability) before a fix wins. Each fix lands as a bead in the source repo
- UB_RUNBOOK.md ships at Phase 12 as the source repo's permanent CI gate: CI Miri flags, loom models, fuzz corpora, a SAFETY-comment template, and a minimum Clippy lint group. A cargo-deny template ships alongside
- Non-goals are explicit: not /rust-unsafe-code-exorcist (block-by-block classification), not /deadlock-finder-and-fixer (hangs and deadlocks). Use this when "no UB" is a shipping requirement, not when "remove one unsafe" is
The audit is bucket-first, not block-first. The 25 taxonomy buckets (aliasing, provenance, alignment, validity, uninitialized memory, transmute/union punning, data races, Send/Sync, Pin, FFI contracts, panic safety, std-library trait invariants, refcount lifecycle, async drop, inline assembly, and the rest) each get their own sweeper. Static sweeps surface candidates; dynamic tools prove or dismiss each. A 10-round convergence floor stops false-positive whack-a-mole: the loop runs at least 10 rounds and stops only after two consecutive rounds with fewer than 3 new genuine findings and ev