跳到正文
cccskills
免费在 GitHub 公开

reproduce-and-fix-issues

Use when a trusted bug or performance report needs reproduction and fix. Not for untrusted reports or scope beyond the named feature.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md4.3 KB
  • agents/openai.yaml150 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Reproduce and fix issues

Contract

FieldBound contract
TriggerReproduce and fix a trusted bug or performance report.
AuthorityHuman-gated: asks before credentials, data-at-rest changes, paid actions, publishing, deployment, remote bulk mutation, or irreversible deletion (preview target and consequence); otherwise reversible local: writes only the fix and reproduction evidence; rollback is version control. No remote mutation except the human-approved draft PR.
Side effectDrives the application under test, captures reproduction evidence, and may open one draft PR. No other remote operations.
DoneVerified operational-thread outcome with cleanup.

Refusals

  • Untrusted reports: rejected. The report must name the affected feature, the environment, and the expected versus actual behavior.
  • Scope beyond the named feature: rejected. Do not extend the fix to adjacent features or infrastructure without an explicit new report.
  • Assuming a race condition without at least two independent reproduction attempts: rejected. Fail the reproduction step instead.

Inputs

  • Required: a trusted bug or performance report describing the symptom or regression. The report must name the affected feature, the environment or context, and the expected versus actual behavior.
  • Optional: reproduction steps, a stack trace, a performance profile, or any supplemental diagnostic from the report.

Procedure

  1. Accept the trusted report. Confirm the affected feature, environment, and symptom are identifiable. Stop if the report does not name a feature or an observable failure mode. Done when: the feature, environment, and symptom are confirmed identifiable.
  2. Drive the application in the environment described by the report. Reproduce the exact symptom before proceeding. Done when: the symptom is reproduced.
  3. Capture reproduction evidence: console output, error messages, stack traces, timing data, or screenshots that record the failure. Fail the reproduction step if the symptom cannot be reproduced after at least two independent attempts; do not assume a race condition. Done when: evidence is captured or no-repro is returned.
  4. Identify the root cause within the affected feature's code. Trace the failure to the first invariant violation or unexpected state. Do not widen scope beyond the named feature. Done when: the root cause is identified.
  5. Implement the minimal fix that resolves the root cause. Validate that the fix does not introduce a new failure in the surrounding behavior. Done when: the fix is implemented and surrounding behavior is validated.
  6. Verify the fix resolves the reproduction case: repeat the reproduction steps and confirm the symptom no longer occurs. Done when: the symptom no longer occurs on repetition.
  7. Clean up any temporary artifacts created during reproduction or diagnosis. If the fix is verified and human approval is given, open one draft PR against the relevant branch. If approval is withheld, retain the verified fix locally and stop without publishing. Done when: cleanup is done and the PR is opened or the fix is retained locally.

Failure and recovery

  • no-repro: the symptom cannot be reproduced after at least two attempts. Stop. Return the evidence of the reproduction attempts and a blocked result.
  • non-converged: the fix is implemented but the symptom persists after verification. Do not widen scope. Return the non-converged result with the last verified state.
  • scope-widening-blocked: a root cause lies outside the named feature. Stop. Do not extend the fix without an explicit new report.
  • Rollback: if the fix introduces a regression, revert to the last known-good state using version control. Do not commit the regression.
  • Partial-result rule: if any step stops for a named failure class, return the result at that step with the evidence collected so far. Do not fabricate or assume subsequent steps succeeded.

Output

A verified fix with captured evidence and cleanup, or a blocked/non-converged result with evidence at the stopping step, and if a draft PR is opened the PR URL and change summary.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Use when explaining System V AMD64, ARM AAPCS, RISC-V psABI, stack frames, variadic calls, or FFI register rules. Not for the Rust FFI binding layer: use rust-ffi.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when configuring ADC sampling time, DMA-driven ADC, calibration, or DAC channel setup on bare-metal MCUs. Not for the DMA stream itself: use dma-baremetal.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

af-xdp

免费

Use when creating AF_XDP sockets, configuring UMEM and XSK rings, writing an XDP redirect program, or choosing copy versus zero-copy mode. Not for full kernel bypass: use dpdk.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when a completed session needs an agent-environment retrospective. Not for an engineering retrospective from telemetry: use engineering-retrospective.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when a redacted, trimmed agent transcript must be appended to a GitHub PR or issue body, with human approval and preview. Not for automated or model-initiated insertion.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

agents-md

免费

Use when a repo needs agent setup, AGENTS.md added or made lean, CLAUDE.md audited, or agent instructions scored or pruned. Not for remote, credential, publish, deploy, or irreversible changes.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

查看 OutlineDriven 的全部技能

报告此技能的问题