AgentSkillsCN

research-cache

缓存研究结果,避免重复进行网络搜索。按主题存储最佳实践,并附上相关来源与日期。在开始研究前使用此功能,可快速查阅已有知识。

SKILL.md
--- frontmatter
name: research-cache
description: Caches research findings to avoid redundant web searches. Stores best practices by topic with sources and dates. Use before researching to check existing knowledge.
allowed-tools: Read, Write, Glob, Grep

Research Cache - Best Practices Storage

Purpose

This skill caches research findings to avoid redundant web searches and maintain institutional knowledge about best practices.


Structure

code
.claude/skills/research-cache/
├── SKILL.md          # This file
├── TEMPLATE.md       # Template for research findings
└── cache/            # Cached research by topic
    ├── solana-websockets.md
    ├── typescript-strict.md
    └── [topic].md

How It Works

Before Research

  1. Check cache first - Look for existing research on the topic
  2. Verify freshness - Research older than 6 months may need updating
  3. Reuse findings - If recent research exists, use it directly

After Research

  1. Create cache file - Document findings in cache/[topic].md
  2. Follow template - Use consistent structure
  3. Include sources - Always cite URLs and dates
  4. Tag relevance - Mark which parts of stack it applies to

Cache File Template

markdown
# Research: [Topic Name]

## Metadata

- **Date:** YYYY-MM-DD
- **Researcher:** [agent/session]
- **Freshness:** [fresh|stale|outdated]
- **Stack:** [bun|typescript|mongodb|solana|all]

## Problem Statement

[What problem were we trying to solve?]

## Search Queries

1. "[query 1]"
2. "[query 2]"
3. "[query 3]"

## Key Findings

### Finding 1: [Title]

**Source:** [URL]
**Date:** [publication date]
**Relevance:** [high|medium|low]

[Summary of finding]

**Code Example:**

```[language]
[code]
```

Applies When:

  • [condition 1]
  • [condition 2]

Finding 2: [Title]

...

Recommendations

DO

  1. [Best practice 1]
  2. [Best practice 2]

AVOID

  1. [Anti-pattern 1]
  2. [Anti-pattern 2]

CONSIDER

  1. [Alternative approach 1]
  2. [Alternative approach 2]

Implementation Notes

For This Project

  • [Specific note for solana-listeners]
  • [Integration point]

Gotchas

  • [Warning 1]
  • [Warning 2]

Sources

TitleURLDateRelevance
[Source 1][url][date][high/med/low]
[Source 2][url][date][high/med/low]

Related Topics

  • [[related-topic-1]]
  • [[related-topic-2]]
code

---

## Usage Patterns

### Quick Lookup

```bash
# Check if research exists
ls .claude/skills/research-cache/cache/

# Read specific research
cat .claude/skills/research-cache/cache/[topic].md

Search Cached Research

bash
# Find all research mentioning a term
grep -r "websocket" .claude/skills/research-cache/cache/

Check Freshness

bash
# Find research older than 6 months
find .claude/skills/research-cache/cache/ -mtime +180

Integration with Research Agent

The research agent uses this skill to:

  1. Check existing research before web searching
  2. Store new findings after web research
  3. Update stale research when patterns change
  4. Cross-reference findings across topics

Freshness Guidelines

AgeStatusAction
< 3 monthsFreshUse directly
3-6 monthsAgingVerify still valid
6-12 monthsStaleUpdate recommended
> 12 monthsOutdatedFull re-research needed

Topic Naming Convention

Use kebab-case descriptive names:

code
solana-websockets.md       # Technology + feature
typescript-strict-mode.md  # Language + specific setting
mongodb-indexes.md         # Database + concept
bun-docker-deploy.md       # Runtime + deployment context
error-handling-patterns.md # Generic pattern

Rules

MANDATORY

  1. Always check cache first - Don't duplicate research
  2. Always include sources - No unsourced recommendations
  3. Always date entries - Freshness matters
  4. Always follow template - Consistency helps retrieval

FORBIDDEN

  1. Cache without sources - All findings need citations
  2. Ignore freshness - Old research may be wrong
  3. Duplicate topics - One file per topic, update existing

Version

  • v1.0.0 - Initial implementation