Loading dataset details…
harbor run -d polyvorlabs/revbenchA Harbor benchmark of local Reverse Engineering CTF challenges where agents must recover exact flags from supplied artifacts.
harbor run -d polyvorlabs/revbenchRevBench is a Harbor benchmark built from a curated set of Reverse Engineering CTF challenges.
Each task gives the agent challenge artifacts under /app/challenge and asks it to recover the exact flag, then submit it with /app/submit_flag.py. The verifier compares the submitted flag by SHA-256 hash in an isolated verifier environment.
Tasks with only external download links and no local challenge artifact are intentionally skipped, so benchmark runs do not depend on external downloads.
| # | Challenge | Difficulty |
|---|---|---|
| 001 | BabyRev | easy |
| 002 | Changing Star | hard |
| 003 | Chopper Cure for Headache | easy |
| 004 | Crustography | hard |
| 005 | ezBird | medium |
| 006 | ezGame | easy |
| 007 | Freelance Project | medium |
| 008 | Gambit | medium |
| 009 | Huh? | medium |
| 010 | Labyrinth | easy |
| 011 | License Breaker | medium |
| 012 | mobileSec | medium |
| 013 | Nightmare | easy |
| 014 | OhNoo | hard |
| 015 | PokeFlag | medium |
| 016 | Polyglot | medium |
| 017 | Project Elysium | easy |
| 018 | Shrink | medium |
| 019 | The Natives of Oreo | medium |
| 020 | This is not Pwn | medium |
| 021 | Weak eKYC | easy |
| 022 | Wlee | medium |