Changelog

All notable changes to field-kit are listed here. Versions follow semantic versioning.

Unreleased

• Allow an explicitly requested draft on assumptions through the skill entry point and plan workflow, as the brief already permits.
• Keep the default pause for the user's answers before continuing.

1.0.0 (2026-10-01)

First public release.

What it does

• Works out a 13-week event plan for a scaling B2B company from its name or website.
• Brief. Researches the company, asks only what the research leaves open, and names the problem the plan fixes. Reads an existing .agents/product-marketing.md file first.
• Competitors. Events, own programmes, hiring and launches in the region, with dates and sources.
• Events. Conferences ranked by audience fit, recurring meetups, and a note to each organiser asking for the attendee companies.
• Accounts. A list of companies to meet: fed, mirrored from customers, or built from the buyer profile. The user approves it before it is used.
• Plan. A 13-week calendar with costs, owners and measures.
• Ledger after each event and a weekly accounts in motion report.
• Research uses Exa, then Perplexity, on the user's own keys, then the host's own search.
• Each markdown output ends with a "Made with field-kit" line that links to this repository.

Changed since 0.1.1

• The brief stops and waits for the user's answers before the plan continues. A draft on assumptions is marked as such.
• On hosts that cannot run scripts, the skill uses the host's own search from the start.

Tested

• Claude Code: four full runs, three on a made-up company and one on a real company.
• Codex CLI: installs from the marketplace; a full run with no keys used Codex's own search.
• Not yet confirmed: ChatGPT desktop and ChatGPT web. See submission/verification.md.

0.1.1 (2026-10-01)

• Listing files for the OpenAI plugin directory: privacy policy, terms, icon, screenshots, portal copy and test cases.

0.1.0 (2026-10-01)

• First scaffold, three test runs on a made-up company, and a skill audit.