菜单
Diagnose and fix bugs, manual test failures, CI failures, and regressions through a disciplined feedback loop. Use when Codex needs to reproduce, minimize, hypothesize, instrument, fix, regression-test, and prepare acceptance evidence.
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
Shared vocabulary and decision process for designing module interfaces and test seams. Use when Codex needs to decide where a seam belongs, judge whether a module is deep or shallow, design an interface that serves slicing and acceptance, or make code easier to test and navigate.
Maintain a repository's domain vocabulary and architecture decision records. Use when Codex needs to sharpen terminology, resolve ambiguous domain concepts, update CONTEXT.md, create ADRs, or support grill-with-docs, triage, TDD, diagnosis, or architecture work with durable domain language.
Clarify a feature, refactor, architecture idea, or product plan against existing code and domain docs. Use when requirements, terminology, constraints, or success signals are unclear before to-prd, to-issues, triage, or implementation, and Codex should update CONTEXT.md or ADRs only when domain terms or decisions crystallize.
Scan a codebase for architecture friction and deepening opportunities. Use when Codex needs to audit old code, find shallow modules, improve testability, identify seams before implementation, or report architecture blockers without directly changing product code.
Move issues, external PRs, new ideas, manual test feedback, existing-codebase audits, and regression reports through workflow triage states. Use when Codex needs to classify work, verify claims, detect already-implemented behavior, ask for missing info, create agent-ready briefs, or route changes before implementation.
Lightweight acceptance gate for completed slices, bug fixes, PRs, or implementation branches. Use when Codex needs to verify that work matches the originating issue or PRD, prove acceptance criteria and regression checks, decide pass/fail/blocked before handoff, or record acceptance evidence in GitHub, a PR, or local markdown.
| name | diagnosing-bugs |
| description | Diagnose and fix bugs, manual test failures, CI failures, and regressions through a disciplined feedback loop. Use when Codex needs to reproduce, minimize, hypothesize, instrument, fix, regression-test, and prepare acceptance evidence. |
Build a feedback loop before changing production code. A bug is not ready to fix until there is a signal that can prove the symptom is gone.
Read docs/agents/issue-tracker.md, docs/agents/domain.md, and docs/agents/verification.md if they exist.
Create the tightest red-capable signal available:
The loop should be:
If no loop can be built, stop and ask for the missing artifact or environment. Do not guess a fix.
Run the loop and confirm it shows the reported symptom. Minimize the repro by removing inputs, steps, config, or data until only load-bearing elements remain.
Capture the exact symptom and the command or steps that produce it.
Generate 3-5 ranked hypotheses. Each must make a falsifiable prediction.
Format:
If <cause>, then <probe or change> should <observable result>.
Use project domain terms and relevant ADR constraints.
Probe one hypothesis at a time. Prefer debuggers or focused inspection. If adding temporary logs, tag them with a unique prefix and remove them before handoff.
For performance regressions, measure first, then change one variable at a time.
Turn the minimized repro into a regression test when there is a correct seam. If there is no correct seam, document that architecture risk and use the best available verification.
Then:
Before declaring done:
If the bug reveals poor seams, hidden coupling, or code that cannot be safely tested, recommend a follow-up architecture issue after the fix is complete.