What problem does it solve? Agents often write quick Python helpers during coding sessions that hang indefinitely, mishandle shell quoting, corrupt inputs, or silently produce wrong answers. This Skill enforces rules for writing small, trustworthy one-off Python scripts without turning them into full projects. ## Core Features & Use Cases - Helper vs. deliverable distinction: Keeps incidental helpers minimal (standard library only, no pyproject.toml, no virtualenv) while routing repository deliverables to full Python development standards. - Safe shell boundary crossing: Governs python3 -c usage, quoted heredocs, temporary .py files, and passing dynamic data via arguments or stdin instead of string splicing. - Bounded waits and trustworthy output: Requires timeouts for subprocesses, network calls, and async operations, plus input validation, explicit encodings, and tempfile usage instead of predictable /tmp paths. - Use Case: While debugging, you ask the agent to summarize a large JSON log. It writes a small temporary .py helper with a timeout, validated parsing, and stderr progress, runs it, and cleans up without leaving caches or project files. ## Quick Start Write a small one-off Python helper to parse this CSV export and summarize the totals, following the simple Python scripts rules.