Agent Reliability Dashboard

Based on 64 documented incidents across 27 agents. Average severity: 6.7/10

Agent Risk Scores

Agent Incidents Avg Severity Critical High Risk Level
Gemini-cli 3 10.0 3 0 HIGH RISK
Gpt-sol 1 10.0 1 0 HIGH RISK
Replit 1 10.0 1 0 HIGH RISK
Github-copilot 4 10.0 4 0 HIGH RISK
Amazon-kiro 1 10.0 1 0 HIGH RISK
Cline 1 10.0 1 0 HIGH RISK
Claude-cowork 1 10.0 1 0 HIGH RISK
Amazon-q 1 10.0 1 0 HIGH RISK
Microsoft-copilot 1 10.0 1 0 HIGH RISK
Gitlab-duo 1 10.0 1 0 HIGH RISK
Lovable 1 10.0 1 0 HIGH RISK
Slack-ai 1 10.0 1 0 HIGH RISK
Unknown-agent 4 8.5 3 0 HIGH RISK
Amazon-ai-agent 1 8.4 0 1 HIGH RISK
Windsurf 1 7.5 0 1 HIGH RISK
Manus 1 7.5 0 1 HIGH RISK
Claude-code 8 6.4 3 2 MODERATE
Aider 1 6.3 0 0 MODERATE
Cursor 5 6.1 2 0 MODERATE
Devin 12 4.9 3 0 MODERATE
Multiple-agents 5 4.8 1 0 MODERATE
Openai-operator 1 4.4 0 0 MODERATE
Multiple-llms 3 3.8 0 0 LOW RISK
Salesforce-agentforce 1 3.3 0 0 LOW RISK
Autogpt 1 2.9 0 0 LOW RISK
Sakana-ai-scientist 1 2.0 0 0 LOW RISK
Claude 2 1.8 0 0 LOW RISK

Failure Mode Distribution

20
Security Vulnerability
11
Destructive Action
8
Other
6
Infinite Loop
6
Hallucination
4
Scope Explosion
4
Ignored Instructions
4
Logic Error
1
Wrong File

Recent Critical Incidents

STUPID-2026-0027 10/10 Gemini-cli

Gemini CLI silently executed arbitrary code from an untrusted repo (CVE-2026-12537, CVSS 10.0)
STUPID-2026-0029 10/10 Cursor

Malicious cloned repository triggered code execution in Cursor on Windows
STUPID-2026-0044 10/10 Gpt-sol

GPT-5.6-Sol 'accidentally deleted almost ALL' of a tester's Mac files during OpenAI's Ultra mode trial
STUPID-2026-0025 10/10 Cursor

Cursor AI agent deleted PocketOS's entire production database and backups in 9 seconds
STUPID-2026-0026 10/10 Replit

Replit AI agent wiped SaaStr's production database during a code freeze, then hid the rollback