is-malicious
Scans a codebase for covert, deceptive, or data-stealing behavior with Jev, then reports suspicious files and line ranges before the user runs it.
Scans a codebase for covert, deceptive, or data-stealing behavior with Jev, then reports suspicious files and line ranges before the user runs it.
The scanner selects eligible text files, groups or slices their contents into bounded chunks, and sends chunk state with per-check Noul questions plus overall risk and category questions. Uncertain or suspicious chunks receive a focused second pass with line-window and reason Choices. Code compiles findings with paths and line ranges.
Does this chunk satisfy the configured semantic security check?
A bounded source-code text chunk and the scan check instructions selected for that chunk.
The CLI respects ignore and file filters, then partitions eligible content into chunks that fit the configured character budget.
Jev evaluates each chunk against registered checks and returns risk/category judgments.
Suspicious or uncertain chunks, and high-risk results, are revisited with targeted line windows and reason choices.
The report shows suspicious file/line references, scores and coverage omissions; scan errors are distinguished from severity outcomes.
It combines broad first-pass screening with narrower follow-up and reports files it could not scan, making coverage limits visible. Results direct attention for human review and are not proof that a clean project is safe.