Public: anyone can read this; the crew can change it.

MCP `backups`: where snapshots go and the last ones taken, for a forge admin #5435

Parked opened this
Closed without merging

I added no new code for this retry. I merged current main into the branch and pushed it so the forge re-runs the check. The backups tool and its change from the first pass are as before.

Notes

stopped parked by the crew deploy ecf382c

opened this

I added no new code for this retry. I merged current main into the branch and pushed it so the forge re-runs the check.

The backups tool and its change from the first pass are as before. They are an admin-only, read-only, no-argument tool that returns the destination, the newest fourteen snapshots and the mirror state. The change includes test/mcp-backups.test.ts, one row in docs/MCP.md, and shorter wording on other tools' descriptions to stay under the tools/list size cap.

The check that parked this was test/asks-one-list.test.ts, "an issue assigned to you is a row that says so, and every door counts it". I couldn't reproduce the failure. That test passed on my branch on its own, three times in a row. It also passed together with test/mcp-backups.test.ts, test/mcp.test.ts and test/mcp-size.test.ts after the merge. Nothing in the tool touches that test's code path (the asks list), so I believe it failed for a reason unrelated to this pull, such as a timing flake. If the check goes red on it again, I'd treat that as a flake on main, not a problem in this pull.

The sandbox has no Bun types, so bun x tsc --noEmit can't check types here. The first-pass type check passed once the types were installed. It was not re-run after this merge.

I also pushed a normal merge rather than a rebase, because force-pushing is refused.

To try it: as an admin, call the backups tool from any MCP client.

🤖 Generated with Claude Code

Closes #5415

commented

I added no new code for this retry. I merged current main into the branch and pushed it so the forge re-runs the check.

The backups tool and its change from the first pass are as before. They are an admin-only, read-only, no-argument tool that returns the destination, the newest fourteen snapshots and the mirror state. The change includes test/mcp-backups.test.ts, one row in docs/MCP.md, and shorter wording on other tools' descriptions to stay under the tools/list size cap.

The check that parked this was test/asks-one-list.test.ts, "an issue assigned to you is a row that says so, and every door counts it". I couldn't reproduce the failure. That test passed on my branch on its own, three times in a row. It also passed together with test/mcp-backups.test.ts, test/mcp.test.ts and test/mcp-size.test.ts after the merge. Nothing in the tool touches that test's code path (the asks list), so I believe it failed for a reason unrelated to this pull, such as a timing flake. If the check goes red on it again, I'd treat that as a flake on main, not a problem in this pull.

The sandbox has no Bun types, so bun x tsc --noEmit can't check types here. The first-pass type check passed once the types were installed. It was not re-run after this merge.

I also pushed a normal merge rather than a rebase, because force-pushing is refused.

To try it: as an admin, call the backups tool from any MCP client.

🤖 Generated with Claude Code

commented

Parked by the crew: the check failed ("boot: a forge addressed by IP hands every app a host that works (issue #269) > the scaffolded app deploys and answers — never the FORGE_APP_URL crash loop") and the decider saw no point in another try

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs