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
- 6
- requests
- 4
- upvotes
- 500
- planned
- 5,029
- shipped
- docs/mcp.md lists a work item's difficulty as `low | medium | high` — `trivial` (the fourth level) is missing on get_work_item, create_work_item and update_work_item
- TypeError: Invalid state: Controller is already closed
- (motir-meta) run.md's vitest guard `pgrep -af 'node.*vitest'` matches every sibling shell that runs the same guard, so "wait until empty" loops across sessions hold each other forever with no vitest running
- PrismaClientKnownRequestError:
- PrismaClientKnownRequestError: Transaction API error: A commit cannot be executed on an expired transaction. The timeout for this transaction was 5000 ms, however 19058 ms passed since the start of t…
- (motir-meta) `MIRROR.map.py --order` is red on `main` at 16 pairs against ORDER_PIN 15 — `phase-author.md` renders "Three sweeps" before "THE SURGICAL TOOLS", the twin composes them the other way round
- (motir-core + motir-marketing) The sandbox guide's VS Code route still goes stale and still loses the sign-in — MOTIR-4970 made only `docker run` disposable, and the page tells a Mac user the container can never sign in to Claude Code
- (motir-core) A `modify` peek's foot line says "only the description and the explanation" whatever the plan changed — a renamed card with a new body reads as if its explanation changed and its title did not
- (motir-core) `filter-builder-matrix.test.ts` goes red whenever the matrix runs across 00:00 UTC — rows seeded "today", relative-date cells queried "tomorrow"
- (motir-core) Can likely-ordering-violation reach an EVIDENCE-phrased post-deploy criterion? — measure it, and the answer may be no
- (motir-core) The item page's activity feed prints raw keys for Work type, Executor and Target repo — "changed the executor" — because the row's field-name map never lists them
- v1 plan-approval answers 500 for a plan with no proposals — PLAN_HAS_NO_PROPOSALS is missing from the v1 domain error map
- (motir-core) A scoped `motir run <story>` never holds a DESIGN child's dependents for its gate, and reads decision holds only once — a gate approved mid-drain releases nothing until the scope is re-run
- (motir-core) After a merge-queue CONFLICT is fixed by a push, the re-asked gate's pull-request row still reads "cannot land" — the approval members read counts a queue exit from the OLD head as standing
- (motir-core) A shape advisory for a BODY-ONLY edit sitting above a field move — the fields and the prose contradict and nothing validates that
- (motir-ai) A global lesson still tells authors to find ordering violations by scanning for merge/release/deploy WORDS, the lexical test MOTIR-5426 measured at about 1 in 9
- (motir-meta) run.md's parent run ends on a design stopper without re-reading its gate — an approval that lands mid-run leaves the released chain unbuilt until someone re-runs
- 41 (motir-core) On the plan-review canvas the breadcrumb runs UNDER the top-right search box — from the third crumb on, "Search this plan" covers the crumbs and swallows their clicks
- (motir-core) A COUNTED claim about a card's own blockers is checkable against the graph — `likely-missing-edge` reads keys, and "all four", "every member", "the siblings this card is blocked_by" carry none
- (motir-ai) The `laidLevel` removal note carries a false count as its REASON — "30+ test files drive `lay` and `propose_node` together" measured 57 files that MENTION both; 7 drive both, and the sweep it justified cost nothing