playbook/outfitter-agents/plugins/outfitter/skills/report-findings/references/source-tiers.md

4.2 KiB

Source Authority Tiers

Comprehensive guide to assessing source credibility and appropriate usage.

Tier 1: Primary Sources (90-100% confidence)

Definitive, authoritative sources from creators or standards bodies.

Types:

  • Official documentation — API references, guides from maintainers
  • Original research — peer-reviewed studies, verified data
  • Direct observation — first-hand evidence, tested behavior
  • Canonical references — specifications, RFCs, standards documents

Use for:

  • Factual claims about behavior
  • API signatures and parameters
  • Performance guarantees
  • Version compatibility statements

Characteristics:

  • Created by authoritative source
  • Regularly maintained
  • Clear versioning
  • Accountable authors

Examples:

  • React documentation from reactjs.org
  • RFC 7231 for HTTP semantics
  • MDN Web Docs for browser APIs
  • TypeScript Handbook from typescriptlang.org

Tier 2: Authoritative Secondary (70-90% confidence)

Expert analysis and recognized publications.

Types:

  • Expert analysis — recognized authorities in field
  • Established publications — reputable sources with editorial standards
  • Official guides — sanctioned tutorials, not canonical reference
  • Conference materials — talks from recognized experts

Use for:

  • Best practices and patterns
  • Architecture recommendations
  • Trade-off analysis
  • Implementation strategies

Characteristics:

  • Author has demonstrated expertise
  • Editorial review process
  • Citations to primary sources
  • Generally current

Examples:

  • Martin Fowler's blog on architecture
  • InfoQ articles with expert authors
  • Conference talks from framework maintainers
  • O'Reilly technical books

Tier 3: Community Sources (50-70% confidence)

Collective wisdom and practical experience.

Types:

  • Community discussions — Stack Overflow, GitHub discussions
  • Individual analysis — technical blogs, personal research
  • Crowd-sourced content — wikis, collaborative documentation
  • Anecdotal evidence — reported experiences, case studies

Use for:

  • Practical workarounds
  • Common pitfalls and gotchas
  • Real-world usage patterns
  • Troubleshooting approaches

Characteristics:

  • May be outdated
  • Quality varies significantly
  • Often context-specific
  • Needs cross-referencing

Examples:

  • Stack Overflow answers (highly voted)
  • GitHub issue discussions
  • Dev.to technical articles
  • Reddit technical discussions

Tier 4: Unverified (0-50% confidence)

Use only as starting points for investigation.

Types:

  • Unattributed content — no clear author or source
  • Outdated material — age unknown or clearly stale
  • Questionable provenance — content farms, SEO-driven sites
  • Unchecked AI content — generated without human verification

Use for:

  • Initial leads only
  • Must verify against higher tiers
  • Never cite directly

Warning signs:

  • No author attribution
  • No dates or version numbers
  • Multiple ads, clickbait titles
  • Generic, shallow content
  • Copied from other sources

Tier Assessment Checklist

When evaluating a source:

Factor Higher Tier Lower Tier
Author Known expert Anonymous/unknown
Publisher Authoritative org Content farm
Date Recent, maintained Old, no updates
Citations Links to sources No references
Depth Detailed, nuanced Surface-level
Accuracy Verifiable claims Unverifiable

Usage Guidelines

For Critical Claims

Require Tier 1 or multiple Tier 2 sources:

  • Security recommendations
  • Performance guarantees
  • Breaking changes
  • Migration paths

For Best Practices

Accept Tier 2, cross-reference with Tier 3:

  • Architecture patterns
  • Code organization
  • Testing strategies
  • Tooling choices

For Troubleshooting

Start with Tier 3, verify against Tier 1-2:

  • Error solutions
  • Workarounds
  • Configuration tips
  • Environment setup

Confidence Adjustments

Factors that increase confidence:

  • Multiple independent sources agree
  • Source is recent and maintained
  • Claims are testable and verified
  • Author has relevant expertise

Factors that decrease confidence:

  • Single source only
  • Source is outdated
  • Claims contradict other sources
  • Author expertise unclear