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
- suggestions
- 4
- votes
- 500
- planifiés
- 4 378
- livrés
- (motir-core) `rate-limit-gate.test.ts` pins the BUDGET but not the WINDOW, so two calls straddling a minute boundary read as under-budget and the assertion gets PROJECT_NOT_FOUND
- Planning bug: `AgentRun` named TWO tables across the open-core boundary — 9.1.7/9.1.8 read motir-ai's metering store as the execution record, so the execution record was never owned
- A client-side navigation to `/planning` intermittently commits NOTHING — the RSC payload and the route's chunks both arrive, and then React renders neither the segment nor its loading boundary
- (motir-meta) Narrow the motir-meta PR-id rule — carry `MOTIR-<n>` when `targetRepo` IS motir-meta, so a card whose deliverable is a prompts change closes on its own mergeSelf-improvement — auto-reported quality bugs
- The board card renders no story points — the DTO carries them, every board mock draws the `.pts` chip, and 4.3.4 shipped `done` claiming the board was wired
- (motir-core) E2E flake, 3rd occurrence — `project-square-flow.spec.ts`'s `rank=popular` navigation stalls the full 180s on bulk-4, on the attempt AND the retry, because the retry re-uses the same degraded webServer
- The docs renderer prints backticks literally when `code` sits inside **bold** — `renderInline` is a flat split, so the bold arm emits its own contents as raw text
- The diagnostics verdict counts cancelled `?_rsc=` prefetches as failed requests — so the failed-request branch shadows the idle-renderer branch that is the whole point of the capture
- The platform-admin mock's inlined Tier-0 block has four stale palette values — a link blue, a yellow tint in both themes, and a dark `--color-warning` override `theme.css` dropped
- MOTIR-2611 broke `acceptance-api-docs.spec.ts` and its own PR could not see it — the sandbox guide's step 2 now has TWO Copy buttons, and the acceptance lane only triggers on `acceptance*.spec.ts` paths
- The acceptance lane still stalls on `/planning`'s first paint — at the 60s budget MOTIR-2506 shipped, so find the stall instead of widening the window again
- reportImplementation writes provenance behind a READ permission — the one exception in the token-permission map
- The acceptance lane's config says it "Runs OFF-CLOUD" and then turns cloud ON — the comment a spec author reads to decide whether their assertion can fail
- The sandbox guide's dev-container step is unfollowable from a cold start — a command palette it never says how to open, a file a GUI file manager refuses to create, and a command that only exists once the folder is already openSelf-improvement — auto-reported quality bugs
- `/api/mcp` carries no rate limiter at all — and `expand_item` submits the same billable job the browser route just had capped
- A credential-free post-publish job must assert the PUBLISHED image's CLI carries every command the shipped docs name — today `sandbox-published` logs in and reads only `--version`Self-improvement — auto-reported quality bugs
- A meta org makes two cross-service calls to answer a question decided by a local column — and a motir-ai outage would fail a settings page that does not need motir-ai
- `notes.html` on `origin/main` has 254 entries and TWO numbered #248 — two sibling lesson PRs merged 3m39s apart, each keeping its own number
- The sandbox guide never tells the reader to `docker pull` — `docker run` reuses a stale `:claude`, so anyone who followed it before 2026-08-05 is still on a CLI with no `motir login` while step 4 tells them to run itSelf-improvement — auto-reported quality bugs
- Mirror the two new gates into `SHARED_PLANNING_RULES` — the released-ref absence check and reproduce-before-you-call-it-broken, worded GENERICALLY because the shipped planner plans other people's productsSelf-improvement — auto-reported quality bugs