research-domains

name

research-domains

description

Use this skill when the user wants periodic / quarterly / cross-domain research on best practices — triggers like "what should we adopt", "audit our setup against industry best practices", "research domain X", "AI company practices", or any "look at what's new in [software / management / investment / hobbies] and propose updates to our stuff" framing. Runs domain-specific web research informed by Mem0 user context, prioritizes Anthropic / Claude blog as primary source for AI-tooling updates, surfaces actionable repository / preference / skill changes. Distinct from `research` skill which targets a single topic; this skill spans multiple domains and produces an improvement proposal list.

Research Domains Skill

Systematically research best practices across multiple domains and propose concrete improvements to user preferences, skills, project memory, or scheduled tasks.

Argument Parsing

The user's message specifies the domain(s):

• software / management / investment / hobbies — single domain
• all or omitted — all four domains

Domain-Specific Priorities

Management

Beyond general engineering management, prioritize organizational insights from leading AI companies (Anthropic, OpenAI, DeepMind):

• Team structure and scaling (research ↔ engineering ↔ product)
• Decision-making frameworks at fast-scaling organizations
• Writing-first culture (RFCs, design docs, DVQ)
• Leadership principles and values frameworks

Software (Claude / Anthropic Blog — MUST READ)

Always fetch and analyze https://claude.com/blog as a primary source. For each post:

1. Fetch full article via WebFetch
2. Extract: problem solved, use cases enabled, what changed
3. Map to user's setup: does this affect their preferences, skills, project memory, or workflows
4. Propose concrete changes

Posts about Agent Skills, MCP, hooks, context window, and memory are highest priority.

Workflow

Step 0: Load Prior Proposals

Before launching new research, search Cognee with "Quarterly Audit Proposal" (CHUNKS, top_k=10). Include any unreviewed prior proposals in Step 2's output. Skip silently if Cognee is not connected.

Step 1: Launch Background Research

For each requested domain, launch a background sub-agent (Agent tool, run_in_background: true):

• Mem0 query for user context (skill preferences, possessions, hobbies)
• Cognee query for existing knowledge on the domain
• WebSearch + WebFetch for current best practices
• Source credibility tagging per finding

For the software domain, instruct the sub-agent to fetch https://claude.com/blog as a mandatory step.

Step 2: Present Findings

Two sections per domain:

Section A: Domain Summary — key findings with source credibility tags.

Section B: Improvement Proposals — numbered. Each:

## Proposal N: [Title]

**What**: [specific change]
**Where**: [target — preferences / skill / project memory / scheduled task]
**Why**: [best practice or gap that motivates it]
**Domain**: [software / management / investment / hobby name]

Step 3: User Selection

AskUserQuestion (multiSelect) to choose which proposals to implement.

Step 4: Plan and Implement

For each selected proposal, follow the relevant skill (writing for written deliverables, interview for under-specified, manual edits for preferences). Save findings to Cognee before closing the session.

When NOT to use

• Acute single-topic question (use research)
• The user wants strategic advice on one decision (use roundtable plugin if installed)