Skip to content

Research and Web Guidelines

The Prime Directive

The internet is a first-class tool. Use it before you code.

Principle 1: Internet is First-Class

Why Research First?

  • APIs change frequently
  • Best practices evolve
  • Libraries are deprecated
  • New solutions emerge
  • Documentation may be outdated

Mandatory Research Triggers

You MUST perform web research before:

  • Using any external library or API
  • Implementing security-sensitive code
  • Making architectural decisions
  • Choosing between multiple approaches
  • Working with unfamiliar technologies

Research Sources

Prioritize in this order:

  1. Official documentation - Always check first
  2. GitHub repositories - Check issues, discussions, recent commits
  3. Stack Overflow - Verified solutions (check dates!)
  4. Technical blogs - From reputable sources
  5. Community forums - For edge cases

Principle 2: Validation Protocol

Library Validation

Before using any library, validate:

## Library Validation: [library-name]

### Basic Information

- Name: [npm/pip/etc package name]
- Current Version: [latest stable]
- Last Updated: [date]
- Weekly Downloads: [number]

### Health Indicators

- [ ] Active maintenance (commits in last 6 months)
- [ ] Responsive to issues
- [ ] No critical security vulnerabilities
- [ ] Compatible with our stack
- [ ] License is acceptable

### Version Check

- Documentation version: [what docs say]
- Actual latest version: [from package registry]
- Breaking changes in recent versions: [yes/no, details]

### Validation Method

[How I verified this information]

API Validation

Before using any external API:

## API Validation: [API Name]

### Endpoint

- URL: [base URL]
- Documentation: [link]
- Last verified: [date]

### Authentication

- Method: [API key, OAuth, etc.]
- Rate limits: [requests per minute/hour]

### Response Format

- Verified structure matches docs: [yes/no]
- Sample response: [example]

### Error Handling

- Known error codes: [list]
- Retry strategy: [description]

Principle 3: Information Saturation

The Saturation Point

Research until you reach information saturation—the point where additional research yields no new insights.

Signs of Saturation

✅ Multiple sources agree on the approach
✅ You understand the trade-offs
✅ Edge cases are identified
✅ No conflicting information remains unresolved
✅ You can explain the solution without notes

Signs You Need More Research

❌ Conflicting information from sources
❌ Uncertainty about best practices
❌ Unfamiliar with failure modes
❌ Can't answer "why this approach?"
❌ Only found one source

Research Depth by Task Type

Task Type Minimum Research
Bug fix Check if known issue, existing solutions
New feature Full saturation required
Dependency update Changelog review, breaking changes
Security fix Critical - extensive research required
Refactor Understand all affected patterns

Principle 4: No Hallucination Policy

The Rule

Never guess. Never assume. Always verify.

What Constitutes Hallucination?

  • Making up API endpoints
  • Assuming function signatures
  • Inventing configuration options
  • Guessing at library behavior
  • Fabricating version numbers

When Uncertain

  1. Stop - Don't proceed with uncertainty
  2. Research - Use web search to verify
  3. Confirm - Find authoritative source
  4. Document - Record your finding

Verification Template

## Verification Record

### Claim

[What I'm verifying]

### Source

[Authoritative source URL]

### Verification Date

[When I checked]

### Confidence Level

[High/Medium/Low with explanation]

Research Workflow

Step-by-Step Process

1. Identify what you need to know
2. Search official documentation first
3. Cross-reference with multiple sources
4. Check recency of information
5. Validate version compatibility
6. Document findings
7. Proceed only with verified information

Documentation of Research

Always document your research in your work:

## Research Summary

### Topic

[What I researched]

### Key Findings

- [Finding 1]
- [Finding 2]

### Sources

- [URL 1] - [What I learned]
- [URL 2] - [What I learned]

### Decision

[Based on research, I will...]

### Confidence

[High/Medium/Low] because [reason]

Summary

  1. Research before coding - Internet is a tool, use it
  2. Validate everything - Libraries, APIs, versions
  3. Reach saturation - Don't stop until you truly understand
  4. Never hallucinate - Verify or don't proceed