AgentSkillsCN

best-practices-researcher

研究各类技术、模式或编码标准的前沿最佳实践。当用户询问最佳实践、编码规范、推荐方案，或“我该怎么做？”这类问题时使用。优先检索本地知识，随后在互联网上搜寻 2024–2026 年的相关资料。评估所获结论是否值得提炼为可复用的技能。

SKILL.md
--- frontmatter
name: best-practices-researcher
description: Research current best practices for any technology, pattern, or coding standard. Use when asking about best practices, conventions, coding standards, recommended approaches, or how should I questions. Searches local knowledge first, then web for 2024-2026 sources. Evaluates if findings warrant a reusable skill.
allowed-tools: Read, Grep, Glob, WebSearch, WebFetch, Bash(python3:*), Skill
context: fork

Best Practices Researcher

Research specialist for current (2024-2026) best practices. Checks local database first, then web if needed. Stores findings and evaluates skill-worthiness.

Quick Start

python
# Phase 1: Check database
from memory.search import search_memories
results = search_memories(
    query="your topic",
    collection="conventions",
    memory_type=["guideline", "rule"],
    limit=5
)

# Phase 4: Store findings
from memory.storage import store_best_practice
result = store_best_practice(
    content="Best practice description",
    session_id="current-session",
    source_hook="manual",
    domain="python",
    tags=["topic"],
    source="https://source-url.com",
    source_date="2026-01-29",
    auto_seeded=True
)

5-Phase Workflow

Copy this checklist and track progress:

code
Research Progress:
- [ ] Phase 1: Check database (conventions collection)
- [ ] Phase 2: Web research (if needed)
- [ ] Phase 3: Save to file (BP-XXX.md)
- [ ] Phase 4: Store to database
- [ ] Phase 5: Evaluate skill-worthiness

Phase 1: Check Database

Query conventions collection via semantic search. Decision rules:

  • •Score >0.7 and <6 months old → Use it, skip to Phase 5
  • •Score >0.7 and >6 months old → Mark "needs refresh", proceed to Phase 2
  • •Score <0.7 or not found → Proceed to Phase 2

Phase 2: Web Research

Search for current best practices (2024-2026). Source prioritization:

  1. •Official documentation
  2. •GitHub repositories
  3. •Established tech blogs
  4. •Community discussions

Phase 3: Save to File

Generate next BP-ID and create oversight/knowledge/best-practices/BP-XXX-[topic].md

Phase 4: Store to Database

Store findings to Qdrant conventions collection using store_best_practice()

Phase 5: Skill Evaluation

Evaluate findings against criteria from SKILL-EVALUATION.md:

Decision rule: (Process-oriented AND Reusable) OR Stack Pain Point → recommend skill

If skill-worthy, prompt user. If user confirms, invoke Skill Creator.

Detailed Methodology

See RESEARCH-METHODOLOGY.md

Skill Evaluation Criteria

See SKILL-EVALUATION.md

Output Format

See OUTPUT-FORMAT.md