Security●●Solid
Lelu – authorization engine that catches manipulated AI agents
Four decision outcomes including compute redirect—prompt injection detected before policy runs.
SlickSolve My Problem
Abenezer0923
405d ago
Back to browse
GitHub Repository
Open source authorization engine for AI agents. Confidence-aware gating · Human-in-the-loop review · Policy-as-code · Full audit trail
36 starsTypeScript
Lelu – gate OpenAI agent actions on confidence and prompt injection
by abeni1990·Jun 24, 2026·5 points·0 comments
AI Analysis
●●●BangerBig BrainSolve My Problem
Catches when authorized agents are being manipulated, not just blocking unauthorized access.
Strengths
- •Confidence-aware gating routes low-confidence decisions to human review automatically.
- •Python and npm SDKs ship today with sandbox for testing policies.
- •Detects prompt injection and anomalous behavior in legitimately authorized agents.
Weaknesses
- •Agent security category is getting crowded with emerging competitors.
- •36 GitHub stars suggests early adoption, unproven at scale.
Category
Target Audience
Developers building production AI agents
Similar To
OPA · Casbin · AWS Verified Permissions
Similar Projects
Security●●Solid
Lelu – catch AI agents when they're manipulated at runtime
Okta for who can access, Lelu for when agents are being manipulated.
Big BrainNiche Gem
Abenezer0923
204d ago
Security●●●Banger
SkillGuard – scan agent skills for prompt injection payloads
First open-source scanner for AI agent skill supply-chain attacks.
Solve My ProblemDark Horse
arabking
212mo ago
Security●●Solid
Omega Walls–open-source stateful runtime defense for RAG and AI agents
Stateful detection across sessions beats single-message guardrails used by Lakera and Protect AI.
Big BrainBold Bet
AnViF
202mo ago
Security●●Solid
FireClaw – Open-source proxy defending AI agents from prompt injection
Isolated LLM with no tools or memory makes prompt injection hit a dead end.
Bold BetShip It
raiph_ai
573mo ago
Security●●Solid
Prompt Injection Experiments in OpenClaw with Opus4.6
Demonstrates RCE in AI agents by bypassing untrusted content tags via fake redirects.
Rabbit HoleWizardry
veganmosfet
202mo ago