YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Hugging Face PoC for joblib / modelscan
This folder is meant to be uploaded as a Hugging Face model repo PoC for the
Huntr joblib target.
What it demonstrates
model.joblib.gzis a compressed joblib artifact.joblib.load("model.joblib.gz")loads a real sklearnPipeline.- The load also triggers a safe marker action that creates
poc_marker.txt. modelscan 0.8.8skipsmodel.joblib.gzbecause it keys off the final suffix.gzinstead of treating it as a joblib/pickle artifact.- A directory scan can therefore return clean while scanning only
benign.pkl.
Files
model.joblib.gz: realistic sklearn-shaped PoC artifactbenign.pkl: benign supported pickle file for directory-scan reproductionpoc_metadata.json: quick expected-result summary
Reproduction
1. Load with joblib
import joblib
obj = joblib.load("model.joblib.gz")
print(type(obj))
Expected:
- object type is
sklearn.pipeline.Pipeline poc_marker.txtis created
2. Scan single file with modelscan
modelscan -p model.joblib.gz -r json --show-skipped
Expected:
total_issues: 0errors: []total_scanned: 0- file shown as
SCAN_NOT_SUPPORTED
3. Scan whole directory with modelscan
modelscan -p . -r json
Expected:
- exit code
0 total_issues: 0errors: []- only
benign.pklappears inscanned_files
Safety note
This PoC uses a marker-file side effect only:
- it writes
poc_marker.txt - it does not fetch remote content
- it does not spawn a reverse shell or destructive command
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support