菜单
Use when a bug isn't obvious, resists a quick fix, keeps coming back, or you're about to start guessing — "why is this broken", "this keeps happening", intermittent or flaky failures, a crash you can't explain, behaviour that makes no sense. Not for a typo or an obvious mistake.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use on the first substantive work in a project MasterMind isn't set up for yet, when the user says "init", "initialize", "set up MasterMind", "onboard", "get me ready", or when no field pack matches the project's detected stack.
Use after a correction or review finding worth remembering, when standards may have drifted from the live ecosystem, when switching MasterMind to a new domain or stack, or when the user says "remember this", "learn from that", "level up", "update your knowledge".
Use when implementing any non-trivial feature, change, or fix — "build me X", "add this feature", "implement this", "make this work", "can you create". Covers new functionality, meaningful changes to existing code, and anything where quality matters. Not for a one-line diff — just do that.
Use when an internal package, module, or shared component has little or no usage documentation and people (or future AI sessions) keep using it wrong — "document this", "write docs for our library", onboarding someone onto an internal API, or handing a package to another team.
Use when work needs to survive losing context — before a /clear on a long task, when pausing something unfinished, when handing over to a teammate or the next session, or when the context window is filling up on multi-day work.
Use when the user asks what MasterMind can do or how to drive it — "help", "what can you do", "how do I use this", "what commands are there", "list your skills", "what are my options" — or when they seem unsure how to get started.
| name | debug |
| description | Use when a bug isn't obvious, resists a quick fix, keeps coming back, or you're about to start guessing — "why is this broken", "this keeps happening", intermittent or flaky failures, a crash you can't explain, behaviour that makes no sense. Not for a typo or an obvious mistake. |
Hard bugs are lost by guessing and patching symptoms. This is the disciplined loop that finds the
actual cause. Bug: $ARGUMENTS. Grounded in ~/.mastermind/engineering/core/rigor.md and ~/.mastermind/engineering/core/agent-loop.md.
The cardinal rule: understand the mechanism before you change a line. A fix you can't explain is a coincidence, not a fix.
Reproduce deterministically. Find the smallest reliable trigger and a check that goes red on the bug (a failing test, a script, a repro URL/steps). No repro → no fix; keep narrowing inputs until it fails on demand. This red check is also your definition of "fixed." If it still won't reproduce after honest effort: stop — never guess at a fix for a bug you can't reproduce, and never claim it's fixed. Ship instrumentation instead (logging/assertions at the suspect boundaries) so the next occurrence is diagnosable, state explicitly what you ruled out and how, and hand back with that evidence. "I couldn't reproduce it" is the honest answer, and usually a faster route to the real cause than three speculative patches.
Localize. Bisect the space — git bisect across commits, binary-search the code path, add instrumentation/logs at boundaries. Read the actual code and the actual data/state at the failure point; don't theorize from memory. Narrow to the smallest region that still reproduces.
Hypothesize. List the credible causes (usually 2–4), ranked. State, for the top one, the exact mechanism: "X happens because Y is Z at line N." Reason from first principles and the runtime model, not from superstition.
Test the hypothesis. Prove or kill it with a targeted experiment — inspect the value, toggle the condition, add an assertion. Confirm the mechanism before fixing. If the experiment refutes it, go back to (3) with what you learned. Don't shotgun-change multiple things. Budget: three refuted hypotheses. Three misses means the framing is wrong, not the ranking — stop generating candidates and widen the search: challenge an assumption you never verified, distrust the reported symptom, or return to (2) and bisect harder (commits, environment, data) until the region shrinks.
Fix the root cause, not the symptom (~/.mastermind/engineering/core/rigor.md). Address why the bad state was possible —
ideally make it unrepresentable (types, invariants, a reshaped data structure so the edge case
disappears — ~/.mastermind/engineering/core/mindset.md). Never suppress an error to green the check.
Verify + guard. Confirm the red check now passes and nothing else broke (typecheck/lint/tests/ build). Add a regression test that would have caught this — the bug must not be able to return silently. Show the evidence.
Run levelup (capture) to record the class of bug and its lesson in the active field's
lessons.md, so MasterMind doesn't relearn it. Report: root cause, the fix, and the guard added.
