This repo generates practical security wordlists in 3 different ways:
wordlists/trusted_resolvers.txt contains DNS resolver IPs that, apparently, can be trusted. It is generated every 8 hours.wordlists/ghdb.json contains the Google Hacking Database in JSON format. It is generated once a week.wordlists/ are web fuzzing, payload, and recon wordlists. They are generated once a week.The recon lists are intentionally tiered instead of merged into one huge file. For example, dns_subdomains_standard.txt is the fast DNS tier, while HTTP routes, real-world directories, headers, parameters, payloads, and usernames each stay in separate files so tools can choose the right amount of noise for the job.
Source URLs, licenses, and preprocessing notes are documented in SOURCES.md. New sources should add coverage that is not already represented here and should avoid broad aggregators that mostly duplicate existing payloads or add low-signal junk.
The web fuzzing/discovery wordlists are generated from custom wordlists located in custom_wordlists/ and from URLs that are indicated in scripts/yaml_lists/lists.yaml.
Feel free to submit PRs with new payloads for the custom lists or new high-signal URLs for the generated wordlists.
164 followers · starred Apr 2023
66 followers · starred Jan 2026
84 followers · starred Jun 2023
This repo generates practical security wordlists in 3 different ways:
wordlists/trusted_resolvers.txt contains DNS resolver IPs that, apparently, can be trusted. It is generated every 8 hours.wordlists/ghdb.json contains the Google Hacking Database in JSON format. It is generated once a week.wordlists/ are web fuzzing, payload, and recon wordlists. They are generated once a week.The recon lists are intentionally tiered instead of merged into one huge file. For example, dns_subdomains_standard.txt is the fast DNS tier, while HTTP routes, real-world directories, headers, parameters, payloads, and usernames each stay in separate files so tools can choose the right amount of noise for the job.
Source URLs, licenses, and preprocessing notes are documented in SOURCES.md. New sources should add coverage that is not already represented here and should avoid broad aggregators that mostly duplicate existing payloads or add low-signal junk.
The web fuzzing/discovery wordlists are generated from custom wordlists located in custom_wordlists/ and from URLs that are indicated in scripts/yaml_lists/lists.yaml.
Feel free to submit PRs with new payloads for the custom lists or new high-signal URLs for the generated wordlists.
164 followers · starred Apr 2023
66 followers · starred Jan 2026
84 followers · starred Jun 2023