bulk-rna-seq-batch-correction-with-combat
Warn
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: Use of unsafe deserialization via
pandas.read_pickle. - Both
SKILL.mdandreference.mdinstruct the user to load datasets usingpd.read_pickle()(e.g.,dataset_1 = pd.read_pickle('data/combat/GSE18520.pickle')). - Python's
picklemodule is inherently insecure and can result in arbitrary code execution when loading untrusted data. A malicious actor providing a crafted pickle file could compromise the system running the agent. - [PROMPT_INJECTION]: Vulnerability to indirect prompt injection through external data ingestion.
- Ingestion points: The skill reads user-provided expression tables and pickled matrices using
pd.read_pickleandpd.read_csvas described inSKILL.mdand demonstrated inreference.md. - Boundary markers: There are no instructions to use delimiters or ignore instructions embedded within the processed data files.
- Capability inventory: The skill includes file-writing capabilities such as
adata.write_h5ad('adata_batch.h5ad', compression='gzip')andraw.to_csv('raw_data.csv'). - Sanitization: The skill lacks any steps for sanitizing or validating the content of the input CSV or Pickle files before they are processed and used for downstream analysis.
Audit Metadata