Research with triangulated evidence to inform decisions. Use before ADRs or evaluations.
Conduct epistemologically rigorous research to inform decisions. Standing on the shoulders of giants, not reinventing wheels.
When to use: Before ADRs, when evaluating competing approaches, entering unfamiliar domains, or resolving parking lot items.
When to skip: Decision is low-stakes and reversible, or prior research exists in work/research/.
Inputs: Clear research question(s), decision context, depth constraint (quick/standard/deep).
| Depth | Time | Sources | Use when |
|---|
| Quick scan | 1-2h | 5-10 | Low-stakes, familiar domains |
| Standard | 4-8h | 15-30 | Most ADRs, technology evaluation |
| Deep dive | 2-5d | 50-100+ | Strategic decisions, unfamiliar domains |
Define: primary question, secondary questions, decision this informs, depth constraint.
Epistemological principles: Seek disconfirming evidence (falsifiability), require 3+ sources per claim (triangulation), primary > secondary > tertiary sources.
Tool selection:
| Tool | Best for |
|---|---|
ddgr "query" | Quick scans, no API key needed |
llm -m perplexity "query" | Deep research with citations |
| WebSearch | Reliable fallback |
Gather sources: academic papers, official docs, GitHub repos (stars/activity), engineering blogs, community discussions.
Per source: type (primary/secondary/tertiary), evidence level, key finding, relevance.
| Evidence level | Criteria |
|---|---|
| Very High | Peer-reviewed, production-proven at scale, >10k stars |
| High | Expert practitioners at established companies, >1k stars |
| Medium | Community-validated, emerging consensus, >100 stars |
| Low | Single source, unvalidated, <100 stars |
Save to work/research/{topic}/sources/evidence-catalog.md.
Per major claim: find 3+ independent confirmations, note consensus vs disagreement, assign confidence (HIGH/MEDIUM/LOW), acknowledge contrary evidence explicitly.
Extract patterns: convergence points, gaps, RaiSE-specific vs general findings.
Produce: recommendation with confidence level, trade-offs, implementation implications, risks.
Connect to governance: create/reference ADR if architectural, update backlog if actionable, update parking lot if deferred.
| Item | Destination |
|---|---|
| Report | work/research/{topic}/{topic}-report.md |
| Evidence catalog | work/research/{topic}/sources/evidence-catalog.md |
| Navigation | work/research/{topic}/README.md |
| Next | ADR, backlog item, or parking lot update |
references/research-prompt-template.mdwork/research/