Statsoft CLI Statistical Software Integration Assistant

Problem: statistical software is hard for agents to call reliably

Legacy analysis work is often trapped inside local tools such as SPSS, Stata, SAS, R, and Mplus. Their CLIs differ, paths are scattered, versions drift, and batch entry points are inconsistent. Statsoft-CLI does not replace the statistical software; it adds a detectable, configurable, and controlled bridge for AI workflows.

How it works: detection, configuration, and guarded execution

• Platform routing: selects entry points by Windows, macOS, Linux, and software availability, hiding incompatible Windows-only tools on non-Windows hosts.
• Clarification gate: classifies the first request as Simple, Complex, or Vague. Clear tool + action requests go straight to detection/configuration; multi-tool or uncertain requests enter guided menu confirmation.
• Default-deny posture: persisting config.json, revealing paths/versions, launching third-party binaries, and compiling/running Stan models are off unless explicitly authorized via STATSOFT_AUTO_WRITE, STATSOFT_CONFIRM, STATSOFT_REVEAL, STATSOFT_VERIFY, or STATSOFT_CMDSTAN_RUN.
• Auditable config writes: writes go through write_config.py, which accepts only the canonical config.json under the skill root, takes a timestamped backup, and atomically replaces the file.

Boundaries and cautions

Best for engineers who already use statistical software and want AI agents to connect legacy scripts, data conversion, or batch runs. It is not meant to treat GUI-only tools such as Minitab, JASP, or jamovi as headless batch engines. Running .sps files, downloading/installing software, and network access are high-risk operations; review scripts, downloads, and generated commands before production use.