Using AI to Automate your Builder Log

Below is the steps to create a build-log workflow for Claude and KitFlight, the iOS build-log app. You type a plans reference such as 34-10 step 1; Claude reads that page out of your own KAI PDF, checks this wiki's page for the section, checks the FAA 51% checklist for the matching row and its evidence requirement, and writes a finished KitFlight entry in the app's own field order. About a minute per entry, with part numbers taken from the plans rather than from memory.

The three files below are what you hand to Claude to build it.

The files

• How-To — the setup guide. Folder layout, reading the KAI PDFs, capturing your app's vocabulary, wiring in the 51% checklist, and a list of KitFlight import/export behavior established by testing.
• SKILL — the skill itself. Becomes SKILL.md.
• Reference — the companion reference: data model, tag scheme, writing conventions, FAA evidence requirements, JSON schema. Becomes KitFlight Reference.md.

Setup

1. Start a Claude session with your plans folder and a KitFlight folder connected — Claude Code or Cowork, either works.
2. Upload all three PDFs and ask Claude to write the SKILL and Reference back out as SKILL.md and KitFlight Reference.md.
3. Work through the How-To with Claude. It will ask for what only you can supply: your KAI filenames, your KitFlight categories, tags, record types and contributors, your title convention, and two or three of your own log entries as voice examples.
4. Save SKILL.md as a skill named kitflight-log. It then triggers on /kitflight-log, or whenever you describe build work.
5. Test it against entries you have already written, and check every part number against the PDF before you trust it.

Step 3 is the whole job. The structure transfers between builders; the vocabulary does not, and a log that reads like someone else's is worse than one that reads like yours.