AI/ML●●Solid
Agent Action Guard – AI agent action safety
HarmActionsEval benchmark proves GPT and Claude fail at blocking harmful tool use.
Solve My ProblemNiche Gem
praneeth-v
202mo ago
Back to browse
GitHub Repository
"I'm sorry, Dave. I'm afraid I can't do that." — Lean command guard for AI coding agents.
7 starsPython
HAL – Harmful Action Limiter: Lean command guard for AI coding agents.
by otherland26·Mar 13, 2026·3 points·9 comments
AI Analysis
●●SolidBig BrainSolve My Problem
Structural command parsing beats regex for catching dangerous agent actions.
Strengths
- •Parses command structure, not strings—commit messages mentioning rm don't false positive
- •YAML rules block --force unless --force-with-lease is present, path-aware rm -rf checks
- •Copilot hook integration means it ships with rules, not just plumbing
Weaknesses
- •Only works with agents supporting hooks—limited to Copilot-compatible tools
- •3 stars and 5 issues suggest early stage, unclear production adoption
Category
Target Audience
Developers using AI coding agents like Cursor, Copilot, or Claude Code
Similar To
Cursor agent safety · Cline permissions · GitHub Copilot hooks
Similar Projects
Developer Tools●●Solid
A context-aware permission guard for Claude Code
Contextual rules beat allow/deny lists—rm -rf __pycache__ is fine, rm ~/.bashrc is nah.
Solve My ProblemNiche Gem
schipperai
127943mo ago
Security●●●Banger
AI Coding Agent Guardrails enforced at runtime
Blocks dangerous AI agent commands like rm -rf before execution in under 2ms.
Solve My ProblemSlick
cavalrytactics
502mo ago
Developer Tools●●●Banger
Link proof assistant Lean to Claude, fix your code's hidden assumptions
Automates Lean formalization to catch silent assumptions standard unit tests miss.
Big BrainWizardry
kurinikku
212mo ago
Developer Tools●●●Banger
Formal – Formal verification for AI-generated code using Lean 4
Lean 4 proofs for AI code correctness—way more rigorous than unit tests.
Big BrainWizardryZero to One
yamafaktory
442mo ago
Developer Tools●●Solid
Spend weekend Claude tokens making your code bullet-proof with Lean
Lean formalization forces explicit types that expose silent failures Python allows.
Big BrainShip It
kurinikku
312mo ago