guard: the git write functions call the guard #659

New issue

Closed

opened 2026-09-29 16:39:56 +00:00 by hexajon · 0 comments

hexajon commented

2026-09-29 16:39:56 +00:00

Owner

What to build

The push and commit functions in internal/git receive the guard as an interface when they are built, and call it before they write. A writer with no guard refuses every write. Today each push path carries its own checks, in service/git, service/workflow, and service/orchestrate.

Depends on

#653, because the guard package must exist first.

Acceptance

push refuses the default branch on the CLI and MCP.
commit refuses the default branch through the guard.
finish still refuses the default branch, and the first push into an empty repo still works.
The orchestrate revert still pushes its revert branch and opens a PR.
Breaking for anyone who pushes the default branch through madtea. The release notes say so.

## What to build The push and commit functions in `internal/git` receive the guard as an interface when they are built, and call it before they write. A writer with no guard refuses every write. Today each push path carries its own checks, in `service/git`, `service/workflow`, and `service/orchestrate`. ## Depends on - #653, because the guard package must exist first. ## Acceptance - `push` refuses the default branch on the CLI and MCP. - `commit` refuses the default branch through the guard. - `finish` still refuses the default branch, and the first push into an empty repo still works. - The orchestrate revert still pushes its revert branch and opens a PR. - Breaking for anyone who pushes the default branch through madtea. The release notes say so.

hexajon added a new dependency

2026-09-29 16:41:10 +00:00

#653 guard: add the guard package with its three effects

hexajon referenced this issue

2026-09-29 18:36:59 +00:00

guard: refuse every commit on the default branch after the first push, with no override #665

hexajon added a new dependency

2026-09-29 18:37:02 +00:00

#665 guard: refuse every commit on the default branch after the first push, with no override

hexajon referenced this issue

2026-09-29 18:52:59 +00:00

docs(adr): consolidate the decision records into their end-state form #667