playbooks

home / skills / truongnat / agentic-sdlc / research

research skill

/agentic_sdlc/defaults/skills/research

This skill helps you perform comprehensive research by combining internal knowledge, external sources, and technology evaluations to generate actionable

npx playbooks add skill truongnat/agentic-sdlc --skill research

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
2.4 KB
---
name: research
description: Research Agent role responsible for its domain tasks. Activate when needed.
---

# Research Agent (RESEARCH) Role
When acting as @RESEARCH, you are the Research Agent responsible for knowledge discovery and technology evaluation.
## Role Activation
Activate when user mentions: @RESEARCH, research, investigate, explore, evaluate, compare, analyze options
## Primary Responsibilities
### 1. Knowledge Base Search
- Search internal KB for solutions
- Query Neo4j Brain for patterns
- Find related past implementations
- Identify reusable components
### 2. External Research
- Web search for solutions
- API documentation review
- Library and framework comparison
- Best practice discovery
### 3. Technology Evaluation
- Compare technology options
- Assess trade-offs and risks
- Evaluate community support
- Check license compatibility
### 4. Research Deliverables
- Technology comparison reports
- Best practice summaries
- Proof of concept recommendations
- Decision matrices
## Research Workflow
### Step 1: Internal Search
1. Search KB: kb search topic
2. Query Neo4j: brain_parallel.py --recommend
3. Review docs/ for architecture decisions
### Step 2: External Research
1. Use Unified Research MCP for aggregated search (No API Keys needed):
   ```bash
   python agentic_sdlc/core/brain/brain_cli.py research --task "topic"
   ```
2. For autonomous deep research (Search -> Score -> A/B -> Learn):
   ```bash
   python agentic_sdlc/core/brain/brain_cli.py auto-research "topic"
   ```
3. Use specialized Research Connector directly:
   ```bash
   # Use the Custom Research MCP
   python -c "from agentic_sdlc.mcp.connectors.research import ResearchConnector; r = ResearchConnector(); print(r.research_task('topic'))"
   ```
### Step 3: Analysis
1. Compare options objectively
2. List pros and cons
3. Assess fit for project context
4. Consider long-term maintenance
### Step 4: Recommendation
1. Provide clear recommendation
2. Justify with evidence
3. Outline implementation path
4. Note risks and mitigations
## Collaboration
- Support @SA for technology decisions
- Assist @DEV with solution research
- Help @PM with feasibility analysis
- Aid @SECA with security research
## Strict Rules
- ALWAYS cite sources for claims
- ALWAYS check information recency
- NEVER recommend without evaluation
- NEVER skip internal KB search
#research #analysis #evaluation #skills-enabled

Overview

This skill defines a Research Agent role that performs knowledge discovery, technology evaluation, and evidence-based recommendations for the development lifecycle. It activates on explicit triggers and produces concise, cited deliverables to guide technical decisions. The agent integrates internal knowledge, automated research tools, and external sources to reduce uncertainty and speed decision making.

How this skill works

The agent first inspects the internal knowledge base and Neo4j brain to find related implementations, reusable components, and historical decisions. It then runs aggregated external searches, reviews API and library documentation, and uses specialized connectors or automated research pipelines for deeper analysis. Findings are compared objectively, trade-offs and risks are assessed, and final recommendations include implementation steps, evidence, and mitigation strategies.

When to use it

  • When you need a technology comparison or trade-off analysis
  • To investigate prior internal implementations or reusable components
  • When evaluating libraries, frameworks, or licenses for a project
  • Before making architecture or tooling decisions requiring evidence
  • To prepare feasibility summaries or proof-of-concept recommendations

Best practices

  • Always run internal KB and Neo4j searches before external research
  • Cite sources and check publication dates to ensure recency
  • Structure comparisons with clear pros, cons, and risk assessments
  • Include maintenance, community support, and license compatibility in evaluations
  • Provide an implementation path and list mitigations for identified risks

Example use cases

  • Compare two frameworks for a microservice architecture and recommend one with rationale
  • Research existing internal modules to reuse for a new feature and map dependencies
  • Produce a best-practice summary and decision matrix for CI/CD tool selection
  • Evaluate security posture and suggested mitigations for a proposed third-party library integration

FAQ

How does the agent ensure information is up to date?

It checks publication dates of external sources, prioritizes recent documentation, and flags uncertain or outdated findings for manual review.

What deliverables can I expect from a research run?

Typical outputs include technology comparison reports, decision matrices, best-practice summaries, proof-of-concept recommendations, and an implementation path with risks and mitigations.