You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The secret detector no longer fires on obvious placeholder values. A
keyword-style hit (Secret Keyword / password_assignment, both advisory-only)
on a clearly-non-credential value -- a test fixture password = "test", a
docstring example SECRET_KEY = "your-secret-key" -- is now dropped, removing
the top secret-scan false positive on the exact pattern test suites use
everywhere. The filter is value-based and security-preserving:
Only the keyword-style kinds are filtered; the deterministic prefixed kinds
(aws_access_key, github_token, stripe_key, private_key, ...) and the
pre-write deny path are never touched.
A hit is dropped only when EVERY assigned string value on the line is a clear
placeholder, so a real secret sharing a line with a placeholder keeps its hit.
A concat-folded / de-escaped hit (a reassembled value the original line can't
show) is never suppressed, keeping the string-concat obfuscation defense
intact.
Placeholder-shaped values (your-..., <...>, ${...}) are dropped only
when their Shannon entropy is low, so a real high-entropy key that merely
starts with your- still fires. Weak-but-plausibly-real credentials
(admin, password123, s3cr3t, changeme, a bare secret/password) are
NOT treated as placeholders -- committing one is a genuine leak worth a nudge.