I’ll build a local Python tool that turns one UTF-8 CSV or flat JSON-array input format into one CSV or JSON output format. It can select, reorder or rename fields, trim specified text, filter by exact values and remove exact duplicate rows when requested. The scope is one set of rules, up to 20 fields, 10,000 records and 10 MiB per input file. JSON fields must contain scalar values, not nested objects or arrays. Existing identifiers and source files are preserved. Invalid records and missing values are reported rather than guessed. You receive the source, configuration, sample output, validation and exception reports, tests and run instructions. API connections, joins, schedules and additional transformations can be scoped separately. To scope the work, provide a non-sensitive or synthetic sample and expected output; input/output formats and CSV delimiter; field names and protected identifiers; exact operations and their order; filter values; missing-value and duplicate policy; and the target operating system and Python version. Confirm the file, row and field limits. I use OpenAI Codex for AI-assisted development. Before processing your brief or source code with that service, I need confirmation that you have the right to share it and permission covering that material. Customer records and source documents remain local unless you separately identify and authorize content for external AI review. Leaving that authorization blank grants no permission to review document or record content with AI. Do not provide credentials or production personal data. The proposed delivery period starts on a date agreed in writing after the required inputs, permissions and platform contracting steps are complete, following the platform’s actual rules. The package includes 1 consolidated written revision within this scope. Defects against the agreed acceptance checks are corrections; new features or changed rules need a separate scope. Briefs, progress updates, revisions and handover are handled in writing.