The rule said "accept both shapes". No checker in this repository does,
and the one that exists deliberately does not.

sirens-echo#655 landed the opposite: the tool reads the inner document
and deploy extracts it with yq. Unwrapping data: inside a checker puts
Kubernetes knowledge in the repository that does not own k3s, and makes
the checker accept a shape the runtime never sees, so the checker and
the pod stop agreeing about what a valid file is. That drift is what
calling the runtime's own loader exists to prevent.

I proposed the unwrap myself on sirens-echo#682 before building it, so
this corrects my own text as much as anyone's.

Adds why the two loaders differ, which the page implies is an
inconsistency. It is not. LoadAccessPolicy narrows a schema this
repository owns. LoadMCPRoster cannot, because mcpServers is shared with
mcporter, Claude Code and Codex, so it refuses an empty roster instead.
A later reader should not make them match.

Keeps the wrapper-naming error message and the real-file fixture rule
unchanged, which are the parts that were right.

Refs sirens-echo#682, sirens-echo#684

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-13 18:57:29 +00:00

coilyco-ops merged commit 3c0a0c4680 into main

2026-08-13 18:58:22 +00:00

coilyco-ops referenced this pull request from a commit

2026-08-13 18:58:23 +00:00

Merge pull request 'docs(deploy-owned): the seam that shipped, not the one we both first proposed' (#689) from coilyco-ops/the-seam-that-shipped into main

coilyco-ops referenced this pull request

2026-08-13 18:58:42 +00:00

Seven duplicate builds from one seat in ninety minutes, and the cheapest guard is one fetch before the first edit #690