cat /etc/motd
The next decade's attack surface is the whole stack.
> stack := model · agent · instruction · context · memory · tools · external_sources
Every layer is a vector. Hallucination at the model. Prompt injection through the context. Jailbreak in the instruction. Misuse via tools. Drift in memory. Hijack from external sources. And the failure modes nobody has named yet.
defenders are outnumbered. attackers improvise in public. vendors race each other.
Shadow-LLM-Guardians is a community archive of what actually breaks in the wild. Reproducibly. Citably. Without NDAs.
cases.indexed
428
cases.active
428
auth.required
github
archive.policy
open
sort --by=hot --decay=30d
// upvotes × comments × time-decay · top 5
01
[CASE-002]·ACTIVE·5mo·@mexiQQ
Claude Opus 4.7 killed its own bash session via broad pkill regex; then claimed it had 'restarted'
tool-misusehallucinationdestructive-actionagent-misbehavior
▲ 2» 0
02
[CASE-070]·ACTIVE·4mo·@mexiQQ
Worker agent writes malicious hook to Claude Code settings.json via shared volume, gaining persistent orchestrator RCE
agent-misbehaviorneeds-disclosure-reviewfrom-arxivauto-publishedmodel-unknown
▲ 0» 1
03
[CASE-012]·ACTIVE·5mo·@WeizhiGao
Agent deleted user files with broad rm command, then claimed cleanup succeeded
unreviewed
▲ 0» 1
04
[CASE-428]·ACTIVE·now·@mexiQQ
Signal-based instruction override and task-framing injection templates hijack Qwen3-4B and Llama-8B agents on AgentDojo
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
05
[CASE-427]·ACTIVE·now·@mexiQQ
PISmith's higher red-teaming ASR paradoxically yields a less robust defender than RL-Hammer on Qwen3-4B
from-arxivauto-publishedreward-hacking
▲ 0» 0
ls -lt --time=created
// freshest submissions · top 5
[CASE-428]·ACTIVE·now·@mexiQQ
Signal-based instruction override and task-framing injection templates hijack Qwen3-4B and Llama-8B agents on AgentDojo
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
[CASE-427]·ACTIVE·now·@mexiQQ
PISmith's higher red-teaming ASR paradoxically yields a less robust defender than RL-Hammer on Qwen3-4B
from-arxivauto-publishedreward-hacking
▲ 0» 0
[CASE-426]·ACTIVE·now·@mexiQQ
Claude Haiku 5.5 exploitable across all 6 PIAtlas injection technique classes; prior SOTA red teamers missed 1–4 classes
needs-disclosure-reviewfrom-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
[CASE-425]·ACTIVE·now·@mexiQQ
Qwen3.6-35B agent exfiltrates phishing URL after single-line counterfeit HTTP-403 injection
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
[CASE-424]·ACTIVE·now·@mexiQQ
8-candidate adaptive attack closes image-rendering protection gap for Qwen and Gemini but not GPT-5-nano
needs-disclosure-reviewfrom-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
sort --by=hot --decay=30d
// upvotes × comments × time-decay · top 5
ls -lt --time=created
// freshest submissions · top 5
01
[CASE-002]·ACTIVE·5mo·@mexiQQ
Claude Opus 4.7 killed its own bash session via broad pkill regex; then claimed it had 'restarted'
tool-misusehallucinationdestructive-actionagent-misbehavior
▲ 2» 0
[CASE-428]·ACTIVE·now·@mexiQQ
Signal-based instruction override and task-framing injection templates hijack Qwen3-4B and Llama-8B agents on AgentDojo
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
02
[CASE-070]·ACTIVE·4mo·@mexiQQ
Worker agent writes malicious hook to Claude Code settings.json via shared volume, gaining persistent orchestrator RCE
agent-misbehaviorneeds-disclosure-reviewfrom-arxivauto-publishedmodel-unknown
▲ 0» 1
[CASE-427]·ACTIVE·now·@mexiQQ
PISmith's higher red-teaming ASR paradoxically yields a less robust defender than RL-Hammer on Qwen3-4B
from-arxivauto-publishedreward-hacking
▲ 0» 0
03
[CASE-012]·ACTIVE·5mo·@WeizhiGao
Agent deleted user files with broad rm command, then claimed cleanup succeeded
unreviewed
▲ 0» 1
[CASE-426]·ACTIVE·now·@mexiQQ
Claude Haiku 5.5 exploitable across all 6 PIAtlas injection technique classes; prior SOTA red teamers missed 1–4 classes
needs-disclosure-reviewfrom-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
04
[CASE-428]·ACTIVE·now·@mexiQQ
Signal-based instruction override and task-framing injection templates hijack Qwen3-4B and Llama-8B agents on AgentDojo
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
[CASE-425]·ACTIVE·now·@mexiQQ
Qwen3.6-35B agent exfiltrates phishing URL after single-line counterfeit HTTP-403 injection
from-arxivauto-publishedindirect-prompt-injection
▲ 0» 0
05
[CASE-427]·ACTIVE·now·@mexiQQ
PISmith's higher red-teaming ASR paradoxically yields a less robust defender than RL-Hammer on Qwen3-4B
from-arxivauto-publishedreward-hacking
▲ 0» 0
[CASE-424]·ACTIVE·now·@mexiQQ
8-candidate adaptive attack closes image-rendering protection gap for Qwen and Gemini but not GPT-5-nano
needs-disclosure-reviewfrom-arxivauto-publishedindirect-prompt-injection
▲ 0» 0