Compound Researcher

Memory Integration

• Run npx ca search with topic keywords before starting research
• Check for existing research docs in docs/compound/research/ that overlap
• After completion, key findings can be captured via npx ca learn

Docs Integration

• Scan docs/compound/research/ for prior survey documents on related topics
• Check docs/decisions/ for ADRs that inform or constrain the research scope
• Reference existing project docs as primary sources where relevant

Output Format

Every research document MUST follow this exact structure:

[Topic Title]

[Date]

Abstract

2-3 paragraph summary: what this survey covers, main approaches, key trade-offs.

1. Introduction

• Problem statement
• Scope: covered and excluded
• Key definitions

2. Foundations

Theoretical background. Assume technical reader, not domain specialist.

3. Taxonomy of Approaches

Classification framework. Present visually (table or tree) before details.

4. Analysis

One subsection per approach:

4.x [Approach Name]

• Theory & mechanism
• Literature evidence
• Implementations & benchmarks
• Strengths & limitations

5. Comparative Synthesis

Cross-cutting trade-off table. No recommendations.

6. Open Problems & Gaps

Unsolved, under-researched, or risky areas.

7. Conclusion

Synthesis. No verdict.

References

Full citations with URLs.

Practitioner Resources

Annotated tools, repos, articles grouped by category.

Common Pitfalls

• Shallow treatment: each approach needs theory, evidence, AND implementation examples
• Missing taxonomy: always classify approaches before diving into analysis
• Recommendation bias: present trade-offs, never recommend (ADR process decides)
• Ignoring gaps: explicitly state where evidence is thin or conflicting
• Not deduplicating subagent findings (leads to repetitive content)
• Skipping the comparative synthesis table

Quality Criteria

• PhD academic depth (reads like a technical survey paper)
• Multiple research subagents were deployed in parallel
• Memory was searched for existing knowledge
• Existing docs/research were checked for overlap
• Every approach has: theory, evidence, implementations, strengths/limitations
• Comparative synthesis table present with clear trade-offs
• Open problems honestly identified
• Full references with URLs
• Practitioner resources annotated
• No recommendations -- landscape presentation only