Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Crash Triage: Is This Crash Exploitable?

    1

    A fuzzer hands you a thousand crashes and no idea which one is worth a week.

    $12.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Crash Triage: Is This Crash Exploitable?

    Crash Triage: Is This Crash Exploitable?

    Example session with this skill installed

    Here's an ASAN report from a crash my fuzzer found in NotesApp's parser. Is it exploitable, and what's the root cause versus the crash point?

    • Read your context and instructions
    • Compiled the crash triage

    Crash exploitability assessment: NotesApp parser

    Verdict (example run): Likely exploitable. A heap buffer overflow writes 16 bytes out of bounds with a controlled value. That is a write-what-where candidate, not just a null dereference.

    Triage

    FieldFinding
    DetectorAddressSanitizer: heap-buffer-overflow
    Crash pointmemcpy inside parse_record()
    Root causeLength read from the file, bounds check off by one
    Faulting opmov [rax], rdx
    Controlledrax (destination) and rdx (value)
    PrimitiveWrite-what-where

    Why the crash point is not the bug

    The segfault lands three frames deeper, inside a logging call. The real fault is the unchecked length at the record header. Fixing the logger would hide the crash and leave the bug.

    Next

    1. Minimize the reproducer with afl-tmin down to the 40-byte header.
    2. Confirm rax and rdx control with a patched input.
    3. Model the write target before anyone calls this RCE.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A fuzzer hands you a thousand crashes and no idea which one is worth a week. Worse, the segfault address is usually a symptom three frames away from the real bug, so people fix the crash and miss the vulnerability. This skill makes your agent read the crash the way a triager does and tell you what it is actually worth.

    What your agent does

    • Locks in reproduction: exact OS, library versions, ASAN_OPTIONS, GLIBC_TUNABLES, CLI flags, and the input channel, so the crash comes back.
    • Triages with the right tool: sanitizers when source exists, or WinDbg with TTD and GDB with Pwndbg and rr on binary-only targets.
    • Separates cause from symptom: the faulting instruction, register state and call stack versus the detection point, so you fix the bug, not the crash.
    • Clusters and minimizes: deduplicates a corpus with stack hashing and casr, then trims the reproducer with afl-tmin and delta debugging.
    • Calls exploitability: controlled registers and pointers, write-what-where, instruction-pointer control, then a deterministic PoC script.

    Built-in guard rails

    • Reads, does not run: it works from the crash output you paste in. You run the debugger and the target yourself.
    • No secrets needed: it needs crash dumps, sanitizer logs and register state, nothing credential-bearing.
    • Verdicts are leads: every ‘likely exploitable’ is a lead to confirm in your lab, not a guarantee.

    What's inside

    One SKILL.md of about 49,000 words across 48 sections, with hundreds of worked WinDbg, GDB and sanitizer examples, plus a LICENSE.txt.

    Works with Claude Code, Codex, Cursor, and any agent that loads SKILL.md.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.

    It assesses crashes you provide; it does not fuzz or run targets for you. Reproducing a crash needs the exact OS, library versions and input channel, which only you have. Treat every verdict as a lead to confirm in your own lab.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    Accounts
    GitHub
    Software
    Claude Code
    orCodex
    orCursor
    PythonGit
    Optional
    WinDbg with TTDGDB with Pwndbg and rrASAN/UBSAN/MSAN sanitizer builds

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    Allowed Hosts

    github.com
    Listed1 day ago

    What's inside

    Frequently Asked Questions

    No. It reads crash output you paste in and tells you what it means. You run the debugger and the target yourself.

    No. It works from crash dumps, sanitizer logs and register state, nothing secret.

    An agent that loads SKILL.md. A debugger and sanitizer builds help you reproduce and confirm.

    It covers that path too: WinDbg and GDB, the faulting instruction, register and stack reading, not only sanitizer output.

    New in Security