What were the 15 blocking spots?

They included files only on the author's machine, stop rules baked into the code from the author's own past instructions, unpinned component versions, and holes in the test procedure itself.

How was the handoff fixed?

Ten gates were made switchable in kit_config.json, and the distributed ZIP defaults to recording instead of stopping, while the author's own setup still stops as before.

What does the pre-sale check require?

It checks whether a clean-environment test for the ZIP's fingerprint has passed end to end, and if not, it says not to ship.

Large Language ModelsAI Coding AssistantsZenn AI/ML6h agoNov 7, 2026, 19:01 JST

Claude Code-built kit: 15 gates block handoff to buyers

Send on LINE

3 Key Points

What happened
After four months and 418 scripts under scripts, handing the Claude Code-built kit to a buyer, 15 spots stopped the first run — mostly gates the AI added from the author's own stop instructions, not bugs.
Why it matters
Each gate was a correct fix for a past failure, but they were all written assuming the author's own machine and workflow, so sharing the kit means counting gates and sorting which are author-only rules.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

The author cannot read code, so the pipeline grew through instructions rather than review. Every time the author said stop, the AI added a gate, and each gate was a valid response to a real past failure. The problem surfaced only at handoff: the gates were written assuming one machine and one way of working.

When the author compared the distributable ZIP with their own machine, 15 spots blocked a buyer's first run. Reinstalling into a clean environment produced even more stops, and the pattern held: most were not bugs but baked-in assumptions about the author's setup.

The remedy keeps the gates for the author while making 10 of them switchable in kit_config.json, with the distributed ZIP set to record rather than stop. A pre-sale check now refuses to ship a ZIP that has not passed a clean-environment test. The author notes that after fixing the 15 spots, new gates kept appearing with each fresh run.

FAQ

What were the 15 blocking spots?: They included files only on the author's machine, stop rules baked into the code from the author's own past instructions, unpinned component versions, and holes in the test procedure itself.
How was the handoff fixed?: Ten gates were made switchable in kit_config.json, and the distributed ZIP defaults to recording instead of stopping, while the author's own setup still stops as before.
What does the pre-sale check require?: It checks whether a clean-environment test for the ZIP's fingerprint has passed end to end, and if not, it says not to ship.

Zenn AI/MLRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.