ccf-rank

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation provides examples for executing local Python and Node.js scripts (scripts/build_ccf_dataset.py and scripts/query_ccf_rank.mjs) to manage and query the ranking dataset.
  • [EXTERNAL_DOWNLOADS]: The data building script depends on the pypdf Python library for PDF text extraction.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: No network operations or access to sensitive system files were detected. All data processing is local to the skill's environment.
  • [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it processes content from external PDF files during the dataset build process. Ingestion points: scripts/build_ccf_dataset.py reads user-provided PDF files. Boundary markers: No explicit delimiters are used in the generated JSON records. Capability inventory: The skill allows local file system writes (database generation) and console output of parsed rankings. Sanitization: Content is normalized and validated against specific regular expressions (rankings, URLs, abbreviations).
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 08:37 AM