niknorf ↗ analysis ↗ GitHub niknorf/mike
A quiet fork of Mike where niknorf is sharpening how the project-chat assistant reasons about legal work and resists prompt injection.
This is niknorf's fork of Mike, and so far it centers on one idea: making the assistant that answers questions about a legal project both a deeper analyst and a harder target. The proposed direction is additive - a new analytical method laid on top of what Mike already does, covering how the assistant reads material, works through a problem, and stays on task when someone tries to steer it off course.
We don't know much about niknorf beyond the handle, and the fork hasn't diverged from upstream Mike in any shipped sense yet - this is a proposal in the open, not a running product with its own identity. If you're curious how it thinks about legal analysis and injection resistance, GitHub is where the actual shape of it lives.
There's no rebrand, no niche pitch, and no separate deployment story to report. What's here is a focused instinct about assistant quality, and it's worth a click if that's the corner of Mike you care about.
What's in it
- Deeper legal analysis Gives the project-chat assistant a fuller method for working through legal material instead of answering off the surface.
- Prompt-injection resistance Aims to keep the assistant on task when a document or user tries to hijack its instructions.
- Additive by design The change layers on top of Mike's existing behavior rather than replacing anything, so it's a low-risk experiment to follow.
Direction
securitychat-ui
Activity
Threads of work (detailed view)
niknorf wants Mike to treat documents as evidence, not orders
A proposed instructions-only change makes Mike's project-chat assistant dig deeper into legal documents and stops adversarial files from hijacking the conversation.
Pull requests (detailed view)
🟢 Open (1)
#191 feat(project-chat): More thorough and reliable project-chat legal document analysis (prompt-only) niknorf · opened 24d ago