diff --git a/.github/ISSUE_TEMPLATE/README.md b/.github/ISSUE_TEMPLATE/README.md new file mode 100644 index 0000000..03b43ae --- /dev/null +++ b/.github/ISSUE_TEMPLATE/README.md @@ -0,0 +1,37 @@ +# Issue Templates — MiniMax Code + +This folder holds structured issue forms. When you open a new issue, GitHub shows +these as selectable templates. Pick the one that matches your case. + +## Available templates + +| Template | When to use | +|---|---| +| **Badcase** | MiniMax Code didn't behave as expected — wrong output, stuck loop, broken automation, bad team routing, habit learned wrong, etc. | + +## How to file a badcase (the fast path) + +1. **Search first** — your issue may already be reported. If you find a match, add a "Me too" with your repro details, don't open a duplicate. +2. **Upgrade** — make sure you're on the latest version. A lot of "bugs" are already fixed. +3. **Pick the "Badcase" template** — fills in a structured form so triage is fast. +4. **Fill in the form, top to bottom** — every field exists because someone on triage has been blocked by its absence. If a field truly doesn't apply, write `N/A`, don't leave it blank. +5. **Strip the noise** — minimum repro wins. "My screen froze" beats "I clicked 47 things then it froze, here's everything I did today". +6. **Logs > screenshots > text** — for visual bugs, a 10-second screen recording beats 5 screenshots. For backend / agent team bugs, paste the relevant log line, not the whole log. +7. **Don't share secrets** — strip API keys, tokens, personal info before submitting. + +## After you submit + +- A bot will auto-label and route the issue. You don't need to do anything. +- Triage usually responds within **3 business days**. S0 / S1 fast-tracked. +- If you can, stick around for follow-up questions — most badcases need a back-and-forth to nail down. + +## Severity cheat sheet + +| Level | Meaning | Examples | +|---|---|---| +| S0 | Crashed / lost data / unusable | App won't launch, schedule silently dropped, habit data wiped | +| S1 | Core feature broken | Agent team routes wrong, automation doesn't run, Feishu sync broken | +| S2 | Works but wrong | Right team picked, but wrong output; habit learned the wrong pattern | +| S3 | Cosmetic / minor | UI glitch, slow but functional, typo in a label | + +When in doubt, pick the higher severity — triage will downgrade if needed, but won't upgrade. diff --git a/.github/ISSUE_TEMPLATE/badcase.yml b/.github/ISSUE_TEMPLATE/badcase.yml new file mode 100644 index 0000000..d2215b9 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/badcase.yml @@ -0,0 +1,194 @@ +name: Badcase +description: Report a case where MiniMax Code didn't behave as expected — wrong output, stuck loop, broken automation, bad team routing, etc. +title: "[Badcase] " +labels: ["badcase", "needs-triage"] +assignees: [] +body: + - type: markdown + attributes: + value: | + Thanks for taking the time to file a badcase. The more structured the info, the faster we can reproduce and fix. + Please fill in everything below. If a section doesn't apply, write `N/A` — don't leave it blank. + + - type: input + id: version + attributes: + label: MiniMax Code version + description: Found in Settings → About. Example: `1.4.2 (build 20260530)` + placeholder: "1.x.x (build YYYYMMDD)" + validations: + required: true + + - type: dropdown + id: platform + attributes: + label: Platform + description: Where are you running MiniMax Code? + options: + - macOS + - Windows + - Linux + - iOS / iPadOS + - Android + - Web + - Other (describe in "Anything else") + validations: + required: true + + - type: input + id: os_version + attributes: + label: OS version + description: macOS example: `15.4 (24E5085f)` · Windows example: `Windows 11 23H2 (build 22631.4317)` + placeholder: "" + validations: + required: true + + - type: dropdown + id: severity + attributes: + label: Severity + description: How bad is the impact on your work? + options: + - S0 — Crashed / lost data / can't use the app + - S1 — Core feature broken (agent team, automation, habit memory) + - S2 — Works but with wrong output / degraded behavior + - S3 — Cosmetic / minor annoyance + validations: + required: true + + - type: checkboxes + id: area + attributes: + label: Which area is affected? + description: Tick all that apply. If "Other", describe below. + options: + - label: Habit memory (remembers your habits) + required: false + - label: Agent team (builds / routes / runs a team) + required: false + - label: Automation / scheduled task + required: false + - label: Skill / tool calling + required: false + - label: Feishu / Lark integration + required: false + - label: UI / display + required: false + - label: Performance (slow, freeze, high CPU/memory) + required: false + - label: Login / account / sync + required: false + - label: Other + required: false + + - type: textarea + id: summary + attributes: + label: One-line summary + description: What went wrong, in one sentence. Becomes the issue title prefix. + placeholder: "Agent team routed my 'check schedule' request to the skill-expert instead of cron-expert" + validations: + required: true + + - type: textarea + id: expected + attributes: + label: Expected behavior + description: What did you expect to happen? Be specific. + placeholder: | + I expected the request "查一下今天飞书日程" to be handled by the Feishu expert and return today's calendar. + validations: + required: true + + - type: textarea + id: actual + attributes: + label: Actual behavior + description: What actually happened? Include the exact wrong output, error message, or screen state. + placeholder: | + It got routed to the skill-expert, which then tried to find a skill for it and failed with "no matching skill". + validations: + required: true + + - type: textarea + id: repro + attributes: + label: Steps to reproduce + description: Numbered list. Minimum to reproduce — strip everything unrelated. + placeholder: | + 1. Open MiniMax Code + 2. Send the message "查一下今天飞书日程" + 3. Observe routing + validations: + required: true + + - type: textarea + id: frequency + attributes: + label: How often does it happen? + description: Always / sometimes / once. If sometimes, what pattern? + placeholder: | + Always for this exact prompt. I haven't tried variations. + validations: + required: true + + - type: textarea + id: context + attributes: + label: Context (habit / team / automation involved) + description: If a habit, agent team member, or scheduled task is involved, name it. Paste the relevant config / prompt if short. + placeholder: | + Agent team: Mavis Code 功能问答天团 (4 experts) + Routed expert: skill-expert + Expected expert: feishu-expert + validations: + required: false + + - type: textarea + id: workaround + attributes: + label: Workaround (if any) + description: If you found a way around it, share it. Often more valuable than the bug report itself. + placeholder: | + None yet. Switching to user-mode chat and manually asking the Feishu expert works. + validations: + required: false + + - type: input + id: logs + attributes: + label: Logs / screenshots / screen recording (paste or attach) + description: | + - Paste short log snippets inline. + - For longer logs, attach files (drag & drop works). + - Screen recording > screenshots > text description, in that order. + placeholder: | + ![screenshot](https://...) + Or attach `app-2026-06-01T21-30.log` below. + validations: + required: false + + - type: checkboxes + id: checks + attributes: + label: Pre-submit checklist + description: Please confirm before submitting — saves everyone time. + options: + - label: I searched existing issues and didn't find a duplicate + required: true + - label: I'm on the latest version + required: true + - label: I removed sensitive info (API keys, tokens, personal data) + required: true + - label: I can reproduce this on a fresh account / clean state + required: false + + - type: textarea + id: anything_else + attributes: + label: Anything else? + description: Anything that didn't fit above, or context that might help. + placeholder: "" + validations: + required: false