sdlc-agentic-pipeline

Warn

Audited by Socket on Sep 23, 2026

5 alerts found:

Anomalyx3Securityx2
AnomalyLOW
SKILL.md

SUSPICIOUS: The skill’s broad SDLC/deployment scope mostly matches its stated purpose, but it enables high-impact autonomous actions and transitive skill installation while deferring critical onboarding, secret, and SSH behavior to unreviewed referenced files. Main concern is workflow authority plus unverifiable downstream setup, not confirmed malware.

Confidence: 81%Severity: 67%
AnomalyLOW
references/agents/devops-agent.md

The fragment describes a legitimate DevOps automation runbook and does not itself show malicious behavior, credential harvesting, obfuscation, or data exfiltration. It does contain high-impact operational actions, including SSH-controlled Docker deployment, destructive container replacement, automatic Terraform apply, and installation of an unpinned latest MCP server. These warrant strict authorization, input validation, version pinning, and rollback testing, but are not sufficient evidence of malware in the displayed content.

Confidence: 94%Severity: 62%
SecurityMEDIUM
references/setup/service-onboarding.md

No confirmed malware is present in the supplied documentation. However, the onboarding process presents a material supply-chain security risk because it automatically executes unpinned remote code and local installer scripts while collecting and storing high-value credentials. Treat the workflow as unsafe until remote content and scripts are reviewed and integrity-pinned, and secrets are handled through a secure secret manager rather than plaintext project files.

Confidence: 94%Severity: 78%
SecurityMEDIUM
references/agents/pm-agent.md

The fragment is a readable project-manager agent configuration, not evident malware. It contains no direct malicious payload or explicit unauthorized data collection. However, it grants extensive filesystem, shell, network, GitHub, Jira, and deployment-related capabilities and references sensitive credential locations. It should be treated as a high-privilege automation configuration and constrained with least-privilege credentials, command/tool allowlists, secret redaction, and mandatory approval before destructive, repository, or deployment actions.

Confidence: 97%Severity: 72%
AnomalyLOW
references/agents/architect-agent.md

The fragment defines an architecture-design subagent and contains no direct evidence of malware or malicious payloads. It does introduce elevated operational risk because the subagent is permitted to execute shell commands and modify or delete files, and it can access external web and Jira services. Those permissions should be least-privilege restricted and approval-gated, especially for bash, deleteFile, external access, and Jira updates.

Confidence: 98%Severity: 56%
Audit Metadata
Analyzed At
Sep 23, 2026, 03:01 AM
Package URL
pkg:socket/skills-sh/codeartsagent%2Fcodeartsskills%2Fsdlc-agentic-pipeline%2F@64de1bf2997f0ba2509c2c9ce6b532431fa15ad4fa478b0add44373e770aca7d