PhantomShift
An AI Red Team Copilot
We’ve been tinkering with this idea for a while and built Phantomshift! We just published a breakdown — with a full demo video — showing how it plays out in the Recon phase of a red team mission. Instead of spitting out canned “try sudo -l” advice, the AI stays in your terminal, tracks what you’ve already done, and nudges you toward the next logical move.
It’s weirdly fun watching it act more like a sparring partner than a chatbot!
I wrote up the story here if you want to check it out: Medium link
Also check out our page, we would love some feedback on PhantomShift with what you would like to see it include: Site link
Curious what you think — where would you want an AI to jump in and save you time during an op?
Hey Indie Hackers,
I’ve been following this community for a while and finally decided to jump in. I’m currently working on PhantomShift, an AI-driven copilot for red teams and pentesters. The idea came from watching how operators really work under pressure — and realizing that most AI tools give generic advice instead of adapting to the flow of an engagement.
With PhantomShift, we’re experimenting with how AI can observe operator actions, track context, and suggest tactical next steps — like a sparring partner rather than a search box. Beyond operations, we also see huge potential for training: helping new operators build decision-making skills in realistic environments.
I’d love to connect with others here who are experimenting with AI in security, training, or high-pressure workflows. Feedback, advice, or just sharing your experiences would be amazing.
Looking forward to learning from this community and sharing the journey as we build. 🚀
-
1
Love the angle, Sophia. Quick question, how do you capture operator context across tools without leaking sensitive data, and how do you measure that a suggested next step actually helps during an engagement?
-
1
Hey Dennis, thank you and great question!!
Right now PhantomShift runs locally, so nothing gets sent out — that way sensitive details (hosts, creds, etc.) stay inside the operator’s environment. It just keeps a lightweight memory of recent commands and outputs to understand context without logging a full transcript.
For measuring usefulness, we’re experimenting in two ways:
Operator feedback loop — the user can accept, reject, or tweak a suggestion, which gives us a direct sense of whether it actually helped.
Engagement outcome alignment — instead of aiming for “perfect” suggestions, the goal is to reduce dead ends and redundant steps, making the operator’s workflow smoother. We’re tracking that through mock runs and training scenarios right now.
It’s still early, but the focus is: keep it safe, keep it contextual, and keep it useful. Curious — if you were using something like this, how would you want to measure whether the suggestions were actually helping?
-
1
Love the local first setup and the accept reject tweak loop. I would measure help with a simple scorecard. time to first foothold, time to first lateral move, steps per successful action, dead end rate, and suggestion acceptance rate. Add a quick 1 to 5 usefulness tap after each run and compare before and after on the same playbook to see real lift.
btw we could partner up. Me and my team are building HustleAdvisor, a social network where entrepreneurs share practical step by step lessons. If you join the waitlist and later post a short write up about building PhantomShift, we will boost it in the main feed so more people see it.
You get: more users
We get: an entrepreneur on board-
1
Awesome that sounds great! I just signed up to be on the waitlist :) Cannot wait for it to be live!
-
1
Hi Sophia, HustleAdvisor is live!
Would love to see a post about PhantomShift on our platform!
https://hustle-advisor.com -
1
Thanks a lot, Sophia!
See you on the launch in 2 days ;)
-
-
-
Comment