moooon
Motir
Vibe the project. Bring an idea — Motir's three AI layers plan it, track it, and ship it, end to end. You're looking at Motir, built in Motir.
- Vibe Project
- Open Source
- AI Agent
- AI Loop
- 5
- 要望
- 4
- 賛成票
- 443
- 計画済み
- 4,442
- リリース済み
- The upload grant did not sign the snapshot's integrity metadata, so every container-built snapshot was UNREADABLE — pointer moves, every later hydrate silently rebuilds from the repo
- Every buildx job red-lights a PR when Docker Hub blips — `setup-buildx-action` pulls `moby/buildkit` anonymously from `registry-1.docker.io`, with no mirror and no retry
- The boot step's two external calls carry NO deadline — `generate-jitconfig` and Fly `provision` can hang an invocation to its `maxDuration` kill, against `docs/jobs.md` rule 3
- CI FLAKE (2nd occurrence) — `releaseCli.test.ts` wraps `npm pack` and three `bash` guards in vitest's DEFAULT 5s timeout, so `CI package` goes red on a loaded runner for diffs that touch no CLI file
- CI FLAKE — `seedHealthyJobSchedules` seeds the every-minute cron's last run with ZERO margin, so a minute boundary between seed and check turns `system.daily-health-check` red at random
- PLANNING BUG — the prose-vs-graph miss claimed a SIXTH card (MOTIR-2011) while its specified detector (MOTIR-1969) sits unscheduled in the BACKLOG: a card authored the same hour as its blocker recorded `relates_to` where `blocked_by` belonged
- A CI job longer than ~5 minutes OUTLIVES its supervising invocation — `maxDuration = 300` against a 3,600s `DEFAULT_JOB_TIMEOUT_MS`, so the boot's `finally` never runs and the intent holds a fleet slot until the reaper
- `ciRunnerBoot`'s supervision re-runs on EVERY durable-replay pass — only the atomic claim stops a second container, and Inngest's reported run outcome is the loser's `already_claimed`
- `main` was left RED on typecheck — MOTIR-1959 added a required `NormalizedRepo.archived` and MOTIR-1996 merged 60s later off a base that predated it; both PRs were green
- Every `system.ci-runner-boot` run is INVISIBLE in the job_run ledger — the event's `workspaceId: ''` trips the workspace FK and the row is silently dropped
- (motir-core) A resolved repo may be ARCHIVED — the repo set validates MEMBERSHIP but never WRITABILITY, so a plan can pin a repo no PR can be opened against
- PLANNING BUG — the fleet shipped code-complete and UNBOOTABLE: §11 deferred the runner image to a card that then shipped without it, and the three env vars the merged adapter demands had no provisioning card at all
- PLANNING BUG — the MOTIR-1918 amendment pass swept three siblings and missed the card that BUILDS the mechanism, then deferred the per-project runner group to a card that had already shipped
- Planning bug — MOTIR-1926 was authored against a repo ARCHIVED a month earlier: file contents were verified at rung 2, the repo's own writability never was
- Planner rule — a card's TARGET REPO is itself a precondition: verify it is WRITABLE, and SEARCH THE PLAN for its name (motir-meta)
- PLANNING BUG — MOTIR-1900's acceptance requirement was written INTO MOTIR-1785's body as prose but never wired as a dependency edge, so the E2E card read `ready` against a surface that is not on main
- PLANNING BUG — MOTIR-711 was re-filed into Story 1775 with its blockers re-pointed but its DESIGN prerequisite left orphaned in Epic 9, so a UI-touching card sat `ready` with no mockup for the surface it renders
- The acceptance-video CI lane cannot build — EMAIL_PROVIDER=file without E2E_PROD_HARNESS kills `next build`
- PLANNING BUG — a story expansion took over MOTIR-1907's primary mechanism 31 minutes after it was written, and nobody swept the donor: two cards owned one write, and the donor read `ready` against a substrate that does not exist
- PLANNING BUG — the always-Motir-owned reversal moved every tenant's repos behind ONE installation without re-checking the mirror, whose tenancy the shipped schema makes single-workspace by construction