Turns unfamiliar data or log files into concise, actionable summaries. The skill profiles file type and shape, estimates row/record counts, infers column/key types, reports missingness and obvious anomalies, and suggests safe next steps for cleaning or further analysis. It prefers statistics over raw values and redacts likely PII unless the user explicitly requests raw samples.
Use when a user provides a local file path or paste and asks to analyze, validate, clean, or summarize the data. Trigger on commands like ‘profile file’, ‘summarize CSV’, ‘inspect JSONL’, or when preparing datasets for modeling or debugging logs.
scripts/ (e.g. data_triage.py) for structured profiles and JSON output.Works with agents that can run local scripts and handle file paths (any code-capable assistant). The skill emphasizes local processing and safety, avoiding uploads of sensitive data.
Data triage skill for profiling CSV/TSV/JSON/JSONL/text log files. Single Python script using only stdlib with built-in PII redaction by default. Script requires a file path argument so it correctly exits with usage error when run without input — not a crash. Clean, well-structured code with type hints, dataclasses, and proper argparse CLI. No security concerns beyond opt-in --unsafe-raw flag.
Solid, practical skill. Uses only Python stdlib which is great for portability. PII redaction by default is a thoughtful security touch. Architecture is decent but could benefit from a references/ directory and more explicit output contracts in SKILL.md.
