ReasonGate- An explainable gate that blocks LLM prompt injection
Details
- External ID
- 48941051
- Source
- HN
- Company
- —
- Product
- ReasonGate- An explainable gate that blocks LLM prompt injection
- Website domain
- github.com
- Launched
- July 16, 2026
- Cohort
- —
- Upvotes
- 7
- Upvotes percentile
- 0.3972520908004779
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Enrichment
- Theme
- ML inference and model optimization
- Vertical
- Security
- Function
- Compliance & governance
- Audience
- Developer
- AI stance
- AI feature
- Project type
- Commercial product
- Normalized one-liner
- prompt injection prevention for llms
- Manually corrected
- False
Could you build this?
Partial Creating a heuristic or basic classifier proxy for LLM inputs is straightforward, but building an explainable, robust gate against sophisticated, multi-modal prompt injection requires specialized adversarial ML research.
What it would actually take: The system requires a low-latency gateway proxy (built on Rust, Go, or Envoy) that intercepts incoming prompts and evaluates them through a cascade of fast token-level heuristic checks, embedding-based intent anomaly detection, and a fine-tuned safety classifier. The hard part is minimizing false positives while reliably catching indirect, encoded, or multi-turn prompt injections without adding noticeable latency to LLM streams. This requires dedicated adversarial machine learning and application security engineering.
Discussion
11 comments analyzed.
Competitors mentioned: Virus checkers/antivirus software, llm-attacks.org
Concerns raised: Cannot provide 100% protection against prompt injection attacks, Regex-based approach easily bypassed with encoding, obfuscation, or non-English languages, Recall only 76-96%, incompatible with high-stakes financial examples, Maintaining exploit database impractical and requires constant updates, Input and outer instructions unified together, making injection impossible to fully block
Feature requests: Make usable as a skill/plugin, Real demo app instead of interactive demo only, Support for multiple languages beyond English
Competitors
Other products that read as similar to this one — 1049 launches clear the similarity bar, closest 8 shown.
Attention rank: #603 of 1050 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 259 days after the earliest competitor.
- Tokensift, an open-sourced token-efficiency linter for LLM prompts · hn · 2026-08-29 · 6 upvotes · similarity 0.62
- Jo · hn · 2026-06-05 · 10 upvotes · similarity 0.61
- LLVM-jutsu: Anti-LLM obfuscation pass · hn · 2025-12-22 · 8 upvotes · similarity 0.60
- Promptloop · hn · 2026-05-29 · 13 upvotes · similarity 0.57
- preflight · github · 2026-09-20 · 76 upvotes · similarity 0.56
- Tiny-vLLM · hn · 2026-05-29 · 205 upvotes · similarity 0.55
- BonzAI · hn · 2026-05-22 · 5 upvotes · similarity 0.55
- x9-gpt · github · 2026-09-11 · 10 upvotes · similarity 0.55
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a compliance & governance tool for Media & entertainment yet.