pandas-pro
Fail
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references a documentation site at https://jeffallan.github.io/claude-skills/skills/data-ml/pandas-pro/ which has been flagged as malicious by automated URL scanners. Although this is a vendor-owned resource, the blacklist status presents a security risk to users following the link.\n- [DYNAMIC_EXECUTION]: The skill encourages the use of pd.eval() and df.eval() in references/performance-optimization.md and JIT compilation via numba.jit in references/aggregation-groupby.md. These features involve runtime evaluation or compilation of code which can be exploited for arbitrary code execution if untrusted data is included in the evaluation strings or functions.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and transform external data files, creating a vulnerability surface where data-resident instructions could influence the agent's behavior.\n
- Ingestion points: The skill utilizes pd.read_csv and pd.read_parquet in references/performance-optimization.md and references/data-cleaning.md to load data into the agent's context.\n
- Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the provided templates.\n
- Capability inventory: The skill enables full data manipulation; agents using this skill often have permissions for file system operations and code execution.\n
- Sanitization: The provided cleaning reference focuses on data structure and types but does not address sanitizing content against malicious instructional patterns.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata