AgentSkillsCN

create-investigation

在某个功能领域内,新增一项探索性调查。当用户说“为 {feature-name} 探究 {topic}”或“创建 {feature-name} 的 {topic} 调查”时使用。该工具会生成一份调查文件,其中包含方法论、权衡考量、对齐清单以及证据收集等板块。

SKILL.md
--- frontmatter
name: create-investigation
description: >
  Add an investigation exploring one approach within a feature area.
  Use when user says "investigate {topic} for {feature-name}" or "create investigation {feature-name} {topic}".
  Creates investigation file with approach, tradeoffs, alignment checklist, and evidence sections.

Create Investigation

Add an investigation exploring one approach within a feature area.

Usage: When user says "investigate {topic} for {feature-name}" or "create investigation {feature-name} {topic}"

Instructions

  1. Verify feature exists: Check docs/features/{feature-name}/readme.md exists

  2. Create investigation file at docs/features/{feature-name}/investigations/{topic}.md:

markdown
# Investigation: {Topic}

**Feature**: {feature-name}
**Status**: In Progress
**Created**: {YYYY-MM-DD}

## Approach
{Describe this approach - what would we build, how would it work?}

## Tradeoffs

| Pros | Cons |
|------|------|
| {benefit} | {drawback} |

## Alignment

- [ ] Follows architectural layering rules
- [ ] Developer Experience (works with minimal setup)
- [ ] Specification compliance (if applicable)
- [ ] Consistent with existing patterns

## Evidence
{Research findings, code exploration, prior art from similar systems, relevant specs}

## Verdict
*Pending evaluation*
  1. Research the approach:

    • Search codebase for related patterns
    • Check existing ADRs for relevant decisions
    • Look for prior art in similar systems if applicable
    • Document findings in Evidence section
  2. Generate alternatives (quint-style): If this is the first investigation for a feature, briefly note 2-3 other approaches worth investigating. These become future investigation candidates.

  3. Update readme.md: Add row to Investigations table with status "In Progress"

Naming Convention

Use kebab-case for investigation topics: eager-loading, strategy-pattern, channel-based