amd/skills serving-llms-on-instinct
Checked the catalogued skill at serving-llms-on-instinct.
D; retained sample includes a high finding · SD-007 · what this rule accepts as a false positive →
D; retained sample includes a high finding · SD-003 · what this rule accepts as a false positive →
B; retained sample includes a medium finding · SD-007 · what this rule accepts as a false positive →
9 suppressed by AI triage
The script performs a legitimate outbound request to a public model hub to fetch metadata, not exfiltration.
Reference to huggingface.co is a benign outbound URL used for data retrieval.
Function _fetch makes a standard HTTP GET to a trusted host, which is expected behavior.
The outbound request is harmless and integral to the script's functionality.
Defines a standard GitHub URL for a public repository; no malicious code or execution detected.
Defines a standard Docker Hub API URL for a public image; no malicious code or execution detected.
Code only provides user instructions to start Docker; no persistence manipulation performed.
Hardcoded path check for presence of a library, not a path traversal.
Hardcoded system path used for a library check; no malicious file access.