doyourmagic

SUSPICIOUS. The core behavior mostly matches the stated purpose of generating workflow skills from external repos, and the documented sandbox/read-only constraints are proportionate. Risk comes from transitive skill installation and the raw GitHub curl-to-bash promote path; these are not outright malicious, but they expand trust beyond the immediate skill and make execution/install provenance weaker than ideal.

No direct malware behavior is evidenced in this provided fragment because it is installer documentation rather than the implementation. However, it documents a high-blast-radius supply-chain mechanism: remote script execution via `curl ... | bash`, optional verification bypass, and persistent modification of global agent configuration under HOME (plus optional overwriting of many global skills via npx). Treat as high security risk until the referenced bootstrap/install scripts and any installed skills are integrity-pinned, provenance-verified, and reviewed.

Đây là một báo cáo audit, không phải payload thực thi. Phần văn bản không cho thấy malware hoặc obfuscation. Tuy nhiên, các lỗi được báo cáo về ghi cấu hình toàn cục, đường dẫn hook/pre-commit sai, glob không bao phủ .llmwiki và validator fail-open tạo rủi ro bảo mật/vận hành đáng kể nếu mô tả là chính xác. Cần xem mã nguồn thực tế của installer, validator và hook để xác nhận mức độ ảnh hưởng.

No explicit malware behaviors (exfiltration/backdoors/credential theft/obfuscation) are evidenced in the provided fragment because it is documentation and does not include the underlying script implementations. The dominant concern is the intentionally enabled high-impact supply-chain execution pathway: remote code is fetched and executed directly (curl|bash) and remote branch content is cloned and executed, with no integrity/pinning guarantees shown here. This means the security posture of the workflow depends entirely on the referenced upstream scripts and transport integrity; those scripts should be pinned and verified (commit pinning, checksum/signature verification) and audited to reduce the risk of compromised upstream content.