Helen detector
Stale Implementation detector
The Stale Implementation detector finds committed execution intent that moved ahead of its likely code implementation. It is a code-and-recency signal, not proof that the code is wrong.
- Method
- Code-symbol alignment with versioned recency evidence
- Implementation policy
- Advisory
What it catches
Helen compares committed execution artifacts with indexed public code symbols. It selects a strong semantic match and compares the artifact’s update time with the exact symbol version’s start time.
A finding appears when intent is at least 21 days newer than the aligned code interface. Weak matches remain quiet because they do not establish an implementation relationship.
Evidence Helen records
The finding names the artifact, selected symbol, repository path, alignment score, artifact update, symbol-version timestamp, and measured lag in days.
Helen abstains when the best score is tied or the exact symbol version is unavailable. Ambiguous evidence cannot safely clear or create a finding.
How to reconcile it
Inspect whether the changed intent requires code work. Update the implementation when it does, link the artifact to the correct symbol when the match is wrong, or dismiss the finding when the change was documentation-only.
Detector boundary
It is advisory and abstains when alignment, symbol version, or score coverage is incomplete or ambiguous.
Research basis
Eick, S. G., Graves, T. L., Karr, A. F., Marron, J. S., & Mockus, A. (2001). Does code decay? Assessing the evidence from change management data. IEEE Transactions on Software Engineering, 27(1), 1–12.