autonomous-loops

Warn

Audited by Socket on Sep 19, 2026

5 alerts found:

Securityx3Anomalyx2
SecurityMEDIUM
payloads.md

The fragment is a high-risk penetration-testing automation toolkit. It contains explicit brute-force credential testing, SSH access attempts, vulnerability scanning, exploit-style HTTP payloads, concurrent network operations, and a sudo fallback. These capabilities could enable unauthorized access or service disruption if misused. No clear malware indicators such as data exfiltration, persistence, reverse shells, cryptomining, destructive system modification, or suspicious command-and-control infrastructure are present in the visible portion. It should not be executed against systems without verified authorization, and its inclusion in a general-purpose package would warrant strong scrutiny.

Confidence: 98%Severity: 90%
SecurityMEDIUM
guides/autonomous-pentest-orchestration-guide.md

The fragment describes a dual-use autonomous penetration-testing framework. It contains high-impact sinks for network scanning, exploit execution, shell and privilege verification, and infrastructure modification, but no visible data theft, persistence, obfuscation, or other clear malware behavior. Safety controls and human approval are present conceptually but appear incomplete and rely on unspecified implementations and uninitialized state. It should be treated as security-sensitive offensive tooling and reviewed before deployment, particularly the arbitrary custom-script execution, scope enforcement, approval handling, and rollback behavior.

Confidence: 96%Severity: 78%
AnomalyLOW
guides/autonomous-decision-making-guide.md

The fragment appears to support an autonomous penetration-testing workflow rather than perform malware behavior. It contains explicit exploitation examples and materially weak command/scope controls: `ALLOWED_TARGETS` is unused, target parsing is incomplete, and forbidden-pattern matching is easily bypassed. The audit logger can write caller-controlled data to a caller-controlled path and may leak sensitive context. No network exfiltration, credential theft, persistence, cryptomining, dynamic code execution, or destructive action is present in this portion. The primary concern is unsafe authorization and auditing design, not demonstrated malicious intent.

Confidence: 93%Severity: 68%
SecurityMEDIUM
SKILL.md

SUSPICIOUS/HIGH-RISK. The skill’s actual footprint is an autonomous offensive-security orchestration guide with exploit, brute-force, SQLi, monitoring, and defense-evasion behavior, which conflicts with its 'non-attack meta skill' framing. No malware payload or installer is shown, but the capability is dangerous, broad, and disproportionate for a generic infrastructure meta-skill.

Confidence: 93%Severity: 88%
AnomalyLOW
guides/feedback-loop-design-guide.md

The fragment performs adaptive network reconnaissance and includes an explicit firewall-evasion/stealth scanning fallback. This is potentially risky and could be misused, but it is not sufficient evidence of malware or unauthorized activity by itself. The fixed predictable temporary file is a separate local security concern. The Python monitoring code appears benign. Review authorization, privilege level, input validation, and temporary-file handling in the surrounding code before use.

Confidence: 93%Severity: 62%
Audit Metadata
Analyzed At
Sep 19, 2026, 04:37 PM
Package URL
pkg:socket/skills-sh/brucesongs%2Fkali-claw%2Fautonomous-loops%2F@54105b34c292084da0aa2837fd72484267c347f91dc3d09fe3b67890a162b7a8