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
- 468
- 計画済み
- 4,333
- リリース済み
- validate_plan is advertised on every pass and wired by NO handler — and augment offers modify_node/remove_node it cannot serve
- A planning job that takes longer than 15 MINUTES throws its work away — the job token is minted at submit and expires before the write-back, and the queue wait counts against it
- 8.20.6 A merge leaves a `deploy` card at In Review with NO note — the artifact-evidence gate throws out of the merge sync, which rethrows every error it cannot classify
- 8.24 (motir-core) A proposed dependency edge never reaches the plan canvas — the review model resolves an `add`'s own refs only, so a `modify`'s `patch.blockedByAdd` draws no arrow to the added card
- (motir-core) A leaked fleet_in_flight_slot defers an UNRELATED file's CI-runner admission — four test files clear it only in beforeEach
- (motir-ai) A merge KILLS the deploy in flight — MOTIR-3106's fix never reached this repo, and the `fly-deploy` job group does not protect a running release
- MCP write tools SILENTLY DROP a key their own schema forbids and report success — a card filed with `description` instead of `descriptionMd` lands with no body
- (motir-ai) The SIGTERM drain MOTIR-3223 shipped never runs — `CMD ["pnpm", "start"]` puts two processes between Fly and the handlers, so every stop waits the full 300 s and then KILLS the job
- childStatusCascadeService exempts `bug` children — delete the carve-out; MOTIR-3229's incident was a parent declared done over unfinished children, not a kind problem
- Planning bug: `run.md` tells every run that a `design/*` / `docs/*` BRANCH PREFIX is what skips the E2E + Vitest lanes — MOTIR-3148 makes that mechanism false
- Planning bug: a surface with a LIFECYCLE was planned one state at a time and nobody owned the state AFTER the terminal transition — four cards each answered it locally and one of them answered with a DELETE
- Planning bug: a re-plan verified every claim it made and still missed a fourth root cause TWELVE LINES ABOVE the code it quoted — nothing audits what a diagnosis OMITS
- (motir-core) The shell's `overflow-hidden` does not clip an `absolute` descendant anchored to the INITIAL containing block — a 1px `sr-only` span lengthens the document by 993px, the whole shell scrolls up, and the empty band `dvh` was supposed to fix is still there
- (motir-ai) `min_machines_running = 2` started nothing — the floor is a STOP-floor, so the job-concurrency ceiling MOTIR-2783 raised is still ONE
- 8.22 (motir-core) The scoped run opens the parent pull request off its t=0 CLAIM — a child filed mid-drain is now refused, but only AFTER the PR exists
- Planning bug: a coverage card enumerated uncovered arms as work items without asking which of them a test can REACH — three of five bullets were dead code
- Planning bug: `run.md` says the design-result publish is "keyed off the `design/MOTIR-<n>-<slug>` branch" — it is keyed off ANY resolvable `MOTIR-<n>` in the branch ref or the PR title, on every PR
- Planning bug: a card's "what must NOT change" clause described a sibling's behaviour the sibling had already reversed
- `stallDiagnosis` measures the worst single GAP, but a debounce window opens at the FIRST event — a slow burst with no slow gap stays silent
- (motir-core) `likely-over-gate-sizing`'s MINUTES arm tests `estimateMinutes` (agent + CI) against a ceiling the gate places on AGENT time alone — so it fires on the right-sized 5-pointer the gate's own table endorses