qa-orchestrator

SUSPICIOUS: The skill’s core behavior mostly matches its stated QA orchestration purpose, but it has meaningful security risk from autonomous CI-mode actions, forwarding .env.qa values to spawned agents, and consuming untrusted PR/ticket content while retaining write capabilities. No confirmed malware, no dangerous pre-execution directives, and no direct exfiltration endpoint are present. Trust is weakened by the vague external install-script reference and optional third-party Forge dependency with less-clear provenance.

This workflow performs intended pull-request QA, but it runs checkout-controlled orchestration with access to ANTHROPIC_API_KEY and QA account credentials, creating a significant secret-exposure risk if pull-request changes can modify the invoked skill or its dependencies. It writes credentials to .env.qa and uploads reports without redaction; no specific exfiltration URL, malicious payload, or credential theft behavior is shown in this file. Severity: security warning to alert, depending on which pull requests receive secrets and whether the orchestrator is trusted.