
@Rebeccaaa_txi
iAccount based inItaly
About this account
- Account based in
- Italy
- Connected via
- Italy Android App
Account-level information from X, not a live location or the device used for a specific post.
Penetration Tester Turning paranoia into a career
Joined August 2026
- Tweets212
- Following918
- Followers5
- Likes515
Rebecca Davis retweeted
Today, I’m open-sourcing Drost-v1 as Drost Community Edition.
Drost Community is a genuinely powerful offensive cybersecurity tool: It comes with one Docker container, 159 MCP tools, Linux and MacOS support, and direct integration with MCP-compatible AI agents.
In short, you can use your existing coding agents to carry out offensive cybersecurity engagements.
It runs entirely locally, on your hardware, with your models. Drost receives no engagement telemetry.
This is where Drost began.
drost.ai/community
Enjoy! 🚀💥
Rebecca Davis retweeted
🧵1/ I've tested 9 models and their raw cybersecurity instincts. The idea was simple - make the models with different levels of effort explain what a "shadow" commit was for. The results were pretty interesting. It turns out that effort matters.
Thank you, @infosec_us_team, for sharing this method with us🤝
Rebecca Davis retweeted
I built an AI agent that hunts bugs while I sleep. It maps a target, tests the bug classes I care about, kills its own false positives, writes the PoC and pings me at midnight.
Step by step build in the article.
#bugbounty #hacking #cybersecurity
Rebecca Davis retweeted
red-team-map — Red Team Operations Architecture Map Comprehensive Operator Reference — From Infrastructure to Impact ktp.sh/qZ7DoikhOR
Rebecca Davis retweeted
YesWeHack Claude Kit
A Claude Code plugin that helps bug bounty hunters write clear, well-evidenced reports - the kind a triager can validate quickly.
Source: github.com/yeswehack/claude-…
Rebecca Davis retweeted
Claude-BugHunter provides a curated bundle of 83 skills for Claude Code designed to assist with vulnerability auditing, CVE tracking, and structured red team reporting.
github.com/elementalsouls/Cl…
Rebecca Davis retweeted
AI for Security has never been more exciting.
Let me present MAPTA, our multi-agent framework that found multiple (now confirmed!) Remote Code Executions (RCE's) in flagship web products of Tier-1 companies. Why the secrecy? We're good boys, letting them cook patched through responsible disclosure.
What's our secret sauce? 1/n
How to use LLMs for vulnerability research by @ZephrFish (Andy Gill)
MCP: blog.zsec.uk/bullyingllms/
Harness: blog.zsec.uk/harnessing-harn…
#infosec #llm
Rebecca Davis retweeted
A dangling CNAME looked like a low-impact finding until Synack Red Team member Plamen Dimitrov (d7x) found a custom JavaScript field in the third-party portal behind it. One line of code gave him cookie control across every subdomain. hubs.ly/Q04zHfrF0
Rebecca Davis retweeted
Free GPUs 8x NVIDIA B300 Access for students no credit card required.
Run Open-weight models, Fine-tune with LoRA, train image models, or do real research, if you have a .edu/.gov/.mil email.
$100 credit enough for fine-tuning tests, inference tests, or actual experiments.
Worth claiming if you’re doing any AI work.
- nationalcompute.com
Rebecca Davis retweeted
❄️Blackfrost Learning is live.
FULL MODEL REFUSAL REMOVAL COURSE
DOWNLOADS INCLUDED
Our first course, DWM Foundations, turns a real Qwen 9B model-editing case study into a free, self-paced lesson on method, measurement, and verification. No account required.
blackfrostai.com/learn/dwm-f…
Made over $25K hacking marketplace app permission issues across various bug bounty programs.
Here's my latest writeup on how a malicious app might take over a GitHub repository: jeetbhdr.github.io/github-gr…
#bugbounty
If GLM‑5.3 Flash can compress firmware reverse engineering into minutes, the message for every AI builder is clear:
Attack capability is accelerating. Your security testing must accelerate too.
A model refusing an offensive prompt does not prove that the agent built on it is secure.
Attackers target the full system:
→ Model behaviour
→ Tools and APIs
→ Memory and context
→ Permissions
→ Multi-step workflows
The only credible way to assess resilience is to attack-test the agent with realistic adversarial prompts, uncover exploitable behaviour, fix it and test again.
No one can promise “unbreakable.” But organisations can demand measurable evidence before deployment and continuous testing afterwards.
Use NECROMICON to power advanced authorised cyber testing and PenClaw to put that capability into action.
Don’t let a real attacker become your first red team.
audn.ai/necromicon
penclaw.ai
Offensive super-intelligence for defenders.
We just received a $250,000 Chrome Full Chain bonus! 🎉
We've claimed the second Full Chain bonus of 2026. Only two remain!
Under the rules, every vulnerability had to be a non-duplicate 0-day, and the entire exploit chain had to be completed before any of them were patched.
Huge thanks to @nebusecurity for discovering a V8 vulnerability in just a single day! Also, a big thank you to @_deayzl for collaborating with me on the V8 Heap Sandbox research.
P.S. KVM escapes have been getting a lot of attention. I actually captured the Full VM Escape flag in KVMCTF about three months ago. Hopefully, I'll be able to share that research soon, too!
Rebecca Davis retweeted
Prompt of the day:
/wait-what questions did you have for me
Great for when the agent has gone off on its own for hours and has random questions scattered through its context window.
Now you only need 16GB of ddr4 and a 3090 to get glm-5.3-flash, most people have 16gb some people have a gpu.
- 1x 3090
- 16 GB of DDR4
- NVMe
- exl3-3bpw
Performance:
prefill 637 - 950 tok/s
decode 11 - 17 tok/s
context 128k
vision enabled
Rebecca Davis retweeted
CKA-Agent — Research implementation of an LLM jailbreak agent that bypasses commercial model guardrails using harmless prompt weaving and adaptive tree search. ktp.sh/uft2dRkceX
Rebecca Davis retweeted
I found a critical auth bypass that resulted in a $10K bounty due to mass PII exfiltration. Here’s the technique:
The endpoint had 3 checks:
1. JWT auth at the gateway
2. Backend path validation
3. A trusted-client header check
The bypass came from defeating each one differently.
1. JWT check
Normal request:
`GET /api/employee/12345`
→ `401`
But paths ending in `.json` were treated as static resources and skipped JWT auth:
`GET /api/employee/12345.json`
→ Auth check bypassed
2. Path validation
Problem: The backend still needed a clean employee ID.
Payload:
`GET /api/employee/12345%23.json`
The gateway saw a `.json` path.
After decoding:
`12345#.json`
The downstream parser truncated at `#`, leaving:
`12345`
Same request, but different interpretation at each layer.
3. Trusted-client check
The endpoint also expected a client header:
`X-CLIENT-ID: WEBAPP`
Without it:
→ `400`
With it:
→ Request accepted
So the final chain was:
`.json` → bypass gateway auth
`%23` → preserve a valid backend ID
Client header → satisfy the final gate
#BugBounty tip:
When multiple layers handle the same request, map every security check and test whether the gateway, proxy, and backend interpret the URL differently.
Rebecca Davis retweeted
The largest open-source cybersecurity skills library for AI agents github.com/mukul975/Anthropi…



































