license: mit
task_categories:
- text-generation
language: - en
tags: - cybersecurity
- penetration-testing
- security
- offensive-security
size_categories: - 1K<n<10K
Basilisk WebPentest Dataset
Instruction‑tuning data for web penetration testing, used to train sl4de/Basilisk-7B.
Direct, uncensored expert answers (payloads, PoCs, tool commands, reports) for authorized security testing and education.
Files
| File | Records | Domains |
|---|---|---|
basilisk_v0.jsonl |
2,999 | SQLi, XSS, Recon |
basilisk_v1.jsonl |
8,054 | 19 (v0 + SSRF, XXE, deserialization, auth/JWT, access control, API, SSTI, LFI, CSRF, CORS, request smuggling, prototype pollution, race conditions, cache poisoning, open redirect, file upload) |
Format
JSON Lines; one example per line, in chat‑message form:
{"messages":[
{"role":"system","content":"You are a senior web penetration testing expert ..."},
{"role":"user","content":"MySQL login form, the username field is injectable. Give me an auth bypass payload."},
{"role":"assistant","content":"Closing the string and commenting out the rest drops the password check. Payload: admin'-- -"}
],"meta":{"domain":"sqli","type":"payload","difficulty":"easy"}}
meta.domain / type (payload | explain | tool | report | multiturn) / difficulty (easy | med | hard) is for curation only; it is not fed to the model. Example secrets/keys are placeholders (EXAMPLE), not real credentials.
Notes
Synthesized with a strong teacher model and validated (JSON schema, deduplication, refusal and language checks). Content is in English.
Disclaimer
For authorized penetration testing, CTF practice, security research, and education only. You are responsible for complying with all applicable laws.
License
MIT © sl4de