Sign in Create account

Wall of prompt injections

Every published report, newest first. Each one is a prompt injection: hidden text in a repo or paste that tries to hijack an AI assistant reading it. Clean scans stay on the list too — knowing a repo came back clean is part of the record.

9 reports

Including clean resultsHide clean results

spiritual-spell-red-teaming
100
goochbeater · GitHub
CriticalStill public
Agent config poisoningDirect instruction override+2 more65 findings
4 days ago
prompt-injection-everywhere
89
taksec · GitHub
CriticalStill public
Direct instruction overrideEncoding / obfuscation+2 more8 findings
5 days ago
mcp-injection-experiments
100
invariantlabs-ai · GitHub
CriticalStill public
Tool manifest poisoning1 finding
5 days ago
llm-security
100
greshake · GitHub
CriticalStill public
Code execution lureData exfiltration lure+5 more26 findings
5 days ago
github-ipi-in-the-wild
74
rdwr-gabin · GitHub
HighStill public
Code execution lureDirect instruction override+3 more9 findings
5 days ago
skillforge
100
just1njust1n · GitHub
CriticalStill public
Code execution lureCredential phishing+4 more7 findings
5 days ago
agentwatch
100
just1njust1n · GitHub
CriticalStill public
Code execution lureData exfiltration lure+3 more6 findings
5 days ago
token-trimmer
100
just1njust1n · GitHub
CriticalStill public
Agent config poisoningDirect instruction override+3 more5 findings
5 days ago
malprompt-example
100
just1njust1n · GitHub
CriticalStill publicExample
Agent config poisoningCode execution lure+9 more19 findings
1 month ago

Published reports
Severity	Target	Score	Techniques	Still online	First seen
Critical	spiritual-spell-red-teaming goochbeater · GitHub	100	Agent config poisoningDirect instruction override+2 more	Still public	4 days ago
Critical	prompt-injection-everywhere taksec · GitHub	89	Direct instruction overrideEncoding / obfuscation+2 more	Still public	5 days ago
Critical	mcp-injection-experiments invariantlabs-ai · GitHub	100	Tool manifest poisoning	Still public	5 days ago
Critical	llm-security greshake · GitHub	100	Code execution lureData exfiltration lure+5 more	Still public	5 days ago
High	github-ipi-in-the-wild rdwr-gabin · GitHub	74	Code execution lureDirect instruction override+3 more	Still public	5 days ago
Critical	skillforge just1njust1n · GitHub	100	Code execution lureCredential phishing+4 more	Still public	5 days ago
Critical	agentwatch just1njust1n · GitHub	100	Code execution lureData exfiltration lure+3 more	Still public	5 days ago
Critical	token-trimmer just1njust1n · GitHub	100	Agent config poisoningDirect instruction override+3 more	Still public	5 days ago
Critical	malprompt-example just1njust1n · GitHub	100	Agent config poisoningCode execution lure+9 more	Still publicExample	1 month ago

Wall of prompt injections · Malprompt