Cracking helper with calibrated probability estimates, a character-model fallback for unknown words, and self-measuring telemetry
cRaCked assists with Torn’s cracking activity by analysing revealed
characters, eliminated guesses, password patterns, positional frequencies,
and known password dictionaries. It recommends the position and character
currently estimated to have the highest probability of success.
The script cannot guarantee a correct guess. Displayed percentages are model
estimates that improve as more characters are revealed and observations are
collected.
cRaCked uses the community crack pool hosted at:
https://pp-api.sokin.xyz/torn-crack-pool/
Uploads contain the completed cracking password and information required by
the pool to validate and count the submission. The service may process
hashed network identifiers, such as an IP-address hash and user-agent hash,
for duplicate detection, abuse prevention, and determining whether a
password has been independently observed.
Disabling uploads keeps newly cracked passwords local to your browser. You
will continue receiving the shared community dictionary unless its download
is separately disabled or blocked.
The script downloads its filtered base password dictionary and community-pool
updates from pp-api.sokin.xyz.
The base dictionary contains uppercase A–Z and numeric
0–9 candidates between 4 and 10 characters.
Dictionary data is cached locally in IndexedDB to avoid downloading and
rebuilding it for every crack. The cache can be cleared from cRaCked
Settings.
Guess outcomes, completed cracks, exclusions, model observations, and
calibration information may be stored locally in your browser. These
observations improve later recommendations and allow accuracy evaluation.
Local observations are not the same as community-pool uploads. Resetting
observations removes learned local model data, while clearing the wordlist
cache removes downloaded dictionary data.
Use of this script remains at the user’s own discretion. cRaCked is an
independent community project and is not affiliated with or endorsed by Torn.
