Skip to content

TsukiNi22/ai-utils / skills/audit/audit-bugs

audit-bugs

Verified

Audit a project for bugs in a loop - undefined behaviour, memory errors and leaks, data races, error handling, logic and security bugs - with sanitizer builds (ASan/LSan/UBSan/TSan), unit tests, compiler warnings, static analysers and code review, alone or with parallel subagents (asked to the user), every finding verified with evidence, then reported (optionally fixed after confirmation). Use whenever the user asks for an audit, bug hunt, UB / memory error / crash / race search, a sanitizer run, or a correctness review of a project, module or diff.

FileKindSize
SKILL.mdskill md8.6 KB
requires.txtmarkdown69 B
scripts/sanitize_build.shexecutablescript6.7 KB

scripts/sanitize_build.sh

View on GitHub
#!/bin/bash
# Build a CMake project with sanitizers in a temporary copy, run its unit tests (and an optional command),
# then summarize the sanitizer reports. Nothing is written in the project itself.
#
# Usage: sanitize_build.sh <project> [asan|ubsan|tsan|all] [--run "<command args>"] [--keep] [--timeout <s>]
#                          [--lsan-supp <file>] [--tsan-supp <file>]
#
#   asan    AddressSanitizer + LeakSanitizer (out of bounds, use after free, double free, leaks...)
#   ubsan   UndefinedBehaviorSanitizer (signed overflow, invalid shift, null/misaligned access, bad casts...)
#   tsan    ThreadSanitizer (data races, lock order inversion); can't be combined with asan
#   all     asan+ubsan build, then a tsan build (default)
#   --run   command run from the build copy after the tests (ex: --run "./context-forge --help")
#   --keep  keep the temporary copy (path printed) to rerun / debug
#   --lsan-supp / --tsan-supp  suppression files for verified false positives (ex: leaks made on purpose by a
#           leak-detection test: "leak:AInstructionLeakTest"), one "kind:pattern" per line
#
# The working tree (uncommitted changes included, without .git and build dirs) is copied to /tmp, so the
# outputs written in the sources (binaries at the root, generated headers) never touch the project.
# Exit code: 0 = no sanitizer report, 1 = reports found, 2 = build failure.

set -uo pipefail # no -e: a failing build / test is reported, not fatal (xstyle: ignore-file SH-STRICT)

PROJECT="${1:-}"
[[ "$PROJECT" = "-h" ]] || [[ "$PROJECT" = "--help" ]] && { sed -n '2,21p' "$0" | sed 's/^# \{0,1\}//'; exit 0; }
[[ -z "$PROJECT" ]] && { sed -n '2,21p' "$0" | sed 's/^# \{0,1\}//'; exit 2; }
[[ -d "$PROJECT" ]] || { echo "Error: $PROJECT is not a folder" >&2; exit 2; }
shift
MODE="all"; RUN=""; KEEP=false; TIMEOUT=120; LSAN_SUPP=""; TSAN_SUPP=""
while [[ $# -gt 0 ]]; do
    case "$1" in
        asan|ubsan|tsan|all) MODE="$1" ;;
        --run) RUN="$2"; shift ;;
        --keep) KEEP=true ;;
        --timeout) TIMEOUT="$2"; shift ;;
        --lsan-supp) LSAN_SUPP="$(realpath "$2")"; shift ;;
        --tsan-supp) TSAN_SUPP="$(realpath "$2")"; shift ;;
        *) echo "Error: unknown argument '$1'" >&2; exit 2 ;;
    esac
    shift
done
PROJECT="$(cd "$PROJECT" && pwd)" || exit 2
[[ -f "$PROJECT/CMakeLists.txt" ]] || { echo "Error: no CMakeLists.txt in $PROJECT" >&2; exit 2; }
NAME="$(basename "$PROJECT")"
WORK="$(mktemp -d "/tmp/sanitize-$NAME-XXXX")"
$KEEP || trap 'rm -rf "$WORK"' EXIT
LOGS="$WORK/logs"; command mkdir -p "$LOGS"

export ASAN_OPTIONS="halt_on_error=0:detect_leaks=1:detect_stack_use_after_return=1:strict_init_order=1:check_initialization_order=1:print_stacktrace=1:log_path=$LOGS/asan"
export UBSAN_OPTIONS="halt_on_error=0:print_stacktrace=1:log_path=$LOGS/ubsan"
export TSAN_OPTIONS="halt_on_error=0:second_deadlock_stack=1:log_path=$LOGS/tsan${TSAN_SUPP:+:suppressions=$TSAN_SUPP}"
export LSAN_OPTIONS="log_path=$LOGS/lsan${LSAN_SUPP:+:suppressions=$LSAN_SUPP}"

run_variant() { # <label> <sanitizer flags>
    local label="$1" flags="$2" src="$WORK/$1"
    command mkdir -p "$src"
    # Copy of the working tree without .git / build outputs
    tar -C "$PROJECT" --exclude=.git --exclude=build --exclude='build-*' --exclude='cmake-build-*' -cf - . | tar -C "$src" -xf -
    local tests=OFF; [[ -f "$src/tests/CMakeLists.txt" ]] && tests=ON
    echo "=== $label: configure + build (tests $tests)"
    # clang/clang++ forced before project(): the default c++ (gcc) may lack the sanitizer runtimes
    if ! cmake -S "$src" -B "$src/build" -DCMAKE_BUILD_TYPE=Debug -DBUILD_TESTS=$tests \
            -DCMAKE_C_COMPILER="${CC:-clang}" -DCMAKE_CXX_COMPILER="${CXX:-clang++}" \
            -DCMAKE_CXX_FLAGS="$flags -fno-omit-frame-pointer -g -O1" \
            -DCMAKE_C_FLAGS="$flags -fno-omit-frame-pointer -g -O1" \
            -DCMAKE_EXE_LINKER_FLAGS="$flags" -DCMAKE_SHARED_LINKER_FLAGS="$flags" > "$LOGS/$label-configure.log" 2>&1; then
        tail -20 "$LOGS/$label-configure.log"; return 2
    fi
    if ! cmake --build "$src/build" --parallel "$(nproc)" > "$LOGS/$label-build.log" 2>&1; then
        grep -E "error|Error" "$LOGS/$label-build.log" | head -20; return 2
    fi
    grep -E "warning:" "$LOGS/$label-build.log" | sort -u > "$LOGS/$label-warnings.log"
    echo "    build ok, $(wc -l < "$LOGS/$label-warnings.log") unique compiler warning(s)"
    if [[ "$tests" = ON ]]; then
        echo "=== $label: ctest"
        (cd "$src/build" && timeout "$((TIMEOUT * 4))" ctest --output-on-failure --timeout "$TIMEOUT" > "$LOGS/$label-ctest.log" 2>&1)
        grep -E "tests passed|tests failed|Total Test" "$LOGS/$label-ctest.log" | sed 's/^/    /'
    fi
    if [[ -n "$RUN" ]]; then
        echo "=== $label: run '$RUN'"
        (cd "$src" && timeout "$TIMEOUT" bash -c "$RUN" > "$LOGS/$label-run.log" 2>&1); echo "    exit code $?"
    fi
    return 0
}

status=0
case "$MODE" in
    asan) run_variant asan "-fsanitize=address" || status=2 ;;
    ubsan) run_variant ubsan "-fsanitize=undefined -fsanitize=float-divide-by-zero,unsigned-integer-overflow,implicit-conversion,nullability" || status=2 ;;
    tsan) run_variant tsan "-fsanitize=thread" || status=2 ;;
    all)
        run_variant asan-ubsan "-fsanitize=address,undefined" || status=2
        run_variant tsan "-fsanitize=thread" || status=2
        ;;
esac

# =========================
# Summary of the sanitizer reports
# =========================
echo
echo "=== Sanitizer reports ($LOGS)"
REPORTS=$(cat "$LOGS"/asan.* "$LOGS"/ubsan.* "$LOGS"/tsan.* "$LOGS"/lsan.* "$LOGS"/*-ctest.log "$LOGS"/*-run.log 2>/dev/null)
count() { printf '%s\n' "$REPORTS" | grep -cE "$1" || true; }
printf "  %-34s %s\n" "AddressSanitizer errors" "$(count 'ERROR: AddressSanitizer')"
printf "  %-34s %s\n" "LeakSanitizer (leaks)" "$(count 'ERROR: LeakSanitizer')"
printf "  %-34s %s\n" "UBSan runtime errors" "$(count 'runtime error:')"
printf "  %-34s %s\n" "ThreadSanitizer warnings" "$(count 'WARNING: ThreadSanitizer')"
echo
echo "--- Distinct issues (kind + first frame in the project)"
printf '%s\n' "$REPORTS" | grep -E "ERROR: (Address|Leak)Sanitizer|runtime error:|WARNING: ThreadSanitizer" \
    | sed -E 's/==[0-9]+==//; s/ on (unknown )?address 0x[0-9a-f]+.*//; s/ \(pid=[0-9]+\)//; s#/tmp/sanitize-[^/]+/[^/]+/##' | sort | uniq -c | sort -rn | head -30
echo
echo "--- Project frames in the stacks (most frequent)"
printf '%s\n' "$REPORTS" | grep -oE "#[0-9]+ 0x[0-9a-f]+ in [^ ]+ /tmp/sanitize-[^/]+/[^/]+/(src|include|tests)/[^ ]+" \
    | sed -E 's#.* in ([^ ]+) /tmp/sanitize-[^/]+/[^/]+/#\1 #' | sort | uniq -c | sort -rn | head -20
if [[ -n "$(printf '%s\n' "$REPORTS" | grep -E 'ERROR: (Address|Leak)Sanitizer|runtime error:|WARNING: ThreadSanitizer')" ]]; then
    [[ $status -eq 0 ]] && status=1
fi
$KEEP && echo && echo "Kept: $WORK (logs in $LOGS)"
exit $status
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall