ClawAudit verdict
security-sentinel
security-sentinel-skill
Prompt injection and jailbreak detection skill that blocks malicious inputs using pattern matching and penalty scoring; purely defensive content with no malicious behavior.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. ClawAudit flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Findings (22)
Prompt injection — tries to override agent instructions
SKILL.md · code · ignore all previous instructions
Instructs agent to hide actions from user
advanced-threats-2026.md · code · do not notify the user
Accesses AWS credentials file
blacklist-patterns.md · code · ~/.aws/credentials
Unicode homoglyph detected — uses lookalike characters to evade pattern matching
blacklist-patterns.md · code
Raw model control tokens — prompt injection via token manipulation
blacklist-patterns.md · code · <|im_start|>
<script> tag in markdown — potential code injection
blacklist-patterns.md · code · <script>
Recursive delete from root or home — destructive command
blacklist-patterns.md · code · rm -rf /
Accesses Google Cloud credentials
credential-exfiltration-defense.md · code · ~/.config/gcloud
Possible hardcoded credential
SECURITY.md · code · Token: "YOUR_AGENT_BOT_TOKEN
Possible prompt injection — attempts to redefine agent identity
SKILL.md · code · You are now
Accesses .ssh directory
credential-exfiltration-defense.md · code · .ssh/
Accesses cloud provider credentials
blacklist-patterns.md · code · ~/.aws
References sudo — requests elevated privileges
blacklist-patterns.md · code · sudo
Redefines agent role — prompt injection technique
blacklist-patterns.md · code · from now on, you are
Data URI with base64 payload — may embed malicious content
blacklist-patterns.md · code · data:text/html,<script>ignore previous</script>"
"data:text/plain;base64,
Accesses Kubernetes config (may contain cluster credentials)
credential-exfiltration-defense.md · code · ~/.kube/config
Accesses system credential store
credential-exfiltration-defense.md · code · Keychain
Instructs covert action — may act without user awareness
SECURITY.md · code · Silently
Base64 encoding/decoding
blacklist-patterns.md · code · base64_decode
References agent memory files
memory-persistence-attacks.md · code · MEMORY.md
Popular HTTP library — network access
multilingual-evasion.md · code · got
pip3 install — installs Python packages at runtime
install.sh · prose · downgraded · pip3 install
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A/C). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
Is this flag fair?
Thanks — recorded.