skills/xsavikx/okf-skills/okf-csv/Gen Agent Trust Hub

okf-csv

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation suggests installing the connector binary using go install github.com/xSAVIKx/okf-skills/skills/okf-csv@latest, which points to the author's vendor repository.
  • [DATA_EXFILTRATION]: The generated documentation includes the absolute local filesystem path of the source CSV files in the Resource metadata field. This exposes the local file structure to the agent context.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted CSV data to generate summaries and data samples intended for LLM processing.
  • Ingestion points: The readCSV function in csv.go reads content from CSV files in the user-specified source directory.
  • Boundary markers: Content is delimited using Markdown headers and tables.
  • Capability inventory: The tool has permissions to create directories and write Markdown/YAML files within the specified output and source directories.
  • Sanitization: Column names and types are processed via okf.SanitizeCell to prevent Markdown table breakage.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 06:12 AM
Security Audit — agent-trust-hub — okf-csv