The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted CSV data from construction software exports, which represents a potential attack surface for indirect prompt injection.
Ingestion points: Data enters the skill context through the read_csv and merge_csvs methods in SKILL.md when processing external files.
Boundary markers: The implementation relies on standard CSV structure parsing; no explicit boundary markers or instructions to ignore embedded prompts are present in the processing logic.
Capability inventory: The skill includes filesystem write capabilities via the split_csv and export_csv methods.
Sanitization: The clean_dataframe method performs basic whitespace stripping and column name normalization (lowercasing, replacing spaces with underscores), but it does not specifically sanitize cell content for potential injection strings.
[EXTERNAL_DOWNLOADS]: The skill requires the pandas and chardet Python packages for data analysis and encoding detection.

csv-handler