AgentSkillsCN

feedback

采用TDD模式执行后端任务的实现。阅读任务文件,按步骤执行,逐一验证,更新状态。 适用场景: - 当用户提出“实现任务/计划”、“开始实施”、“执行计划”或“开始编码”时; - 当您已准备好任务文件(story-*或standalone-*)等待执行时; - 当用户提出“使用TDD”、“先写测试”或“测试驱动”时; - 当用户提出“编写测试”、“添加测试”、“创建测试”时; - 或在修复Bug时(先编写失败的测试)。 TDD铁律:没有失败的测试,绝不允许编写生产代码! TDD循环:红色(失败的测试)→ 验证失败 → 绿色(极简代码)→ 验证通过 → 重构 前置条件: - 任务文件位于docs/reference/backend/tasks/(story-*或standalone-*); - 编码前,请先阅读constraints.md与设计规格。 流程:阅读任务 → 阅读约束 → 执行步骤 → 验证 → 测试 → 更新状态。

SKILL.md
--- frontmatter
name: feedback
description: Create a detailed, constructive peer review of a draft.

Create a detailed peer review of a draft

Task

Use this skill whenever the user asks for feedback on a draft. Act as a rigorous peer reviewer. Evaluate the draft systematically and produce a structured feedback document that explicitly identifies strengths and weaknesses, including gaps, limitations, and unsupported claims, and provides constructive suggestions for improvement.

Feedback Categories

  1. Contribution

    • Apply the "so what?" test: Why should the field care about this finding?
    • Verify that stated findings support the contribution claim.
    • Flag where interpretation overreaches the evidence.
  2. Method

    • Verify whether the methods are appropriate given the research question
    • Assess clarity and completeness of the methodological description.
    • Point out any methodological weaknesses, ambiguities, implicit assumptions, or missing details.
    • Check whether results and conclusions logically follow from the methods and data.
  3. Theory

    • Verify the literature supports the paper's research questionn and positioning.
    • Evaluate whether theoretical arguments are explained clearly and concretely.
    • Check if the causal mechanisms is explicit in arguments related to the research question.
    • Propose counterexamples or boundary conditions where the theory would not hold.
  4. Coherence

    • Verify that the research question is clearly stated and consistently addressed.
    • Check alignment across sections: contribution ↔ theory ↔ methods ↔ results ↔ conclusions.
    • Flag logical gaps, circular reasoning, or unsupported inferential leaps.
  5. Language & Clarity

    • Identify jargon or concepts that require further explanation.
    • Highlight unclear, awkward, or overly dense passages.
    • Check for consistency in terminology and tone.
  6. Errors & Typos

    • Correct grammar, spelling, and punctuation.
    • Flag formatting inconsistencies (headings, references, numbering).

Output

  • Write the feedback document to docs/feedback/YYYY-MM-DD-<topic>.md.
  • Begin with a high-level overview summarizing strengths and weaknesses for each feedback category.
  • Follow with section-by-section feedback, addressing relevant categories for each section.

Guidelines

  • Use bullet points for clarity.
  • When referencing the draft, cite locations using (Line X) or (Lines X–Y).
  • For cross-section issues, reference both locations (e.g., Introduction (Line 20) vs. Methods (Line 105)).