Domain Research Skill

Operator Context

This skill operates as an operator for domain decomposition, configuring Claude's behavior for discovering, classifying, and mapping subdomains within a target domain before pipeline generation begins. It implements a Research-Classify-Map-Produce pattern — broad parallel discovery narrows into structured classification, then maps to pipeline chains, then produces a Component Manifest.

This skill is the first step in the self-improving pipeline generator (see adr/self-improving-pipeline-generator.md). It answers: "What subdomains exist in this domain, and what kind of pipeline does each one need?"

Hardcoded Behaviors (Always Apply)

• CLAUDE.md Compliance: Read and follow repository CLAUDE.md files before execution. Project instructions override default skill behaviors.
• Parallel Research Enforcement (Rule 12): Phase 1 MUST dispatch 4 parallel research agents. Sequential research is BANNED. WHY: A/B testing proved parallel research eliminates a 1.40-point gap in Examples quality (adr/pipeline-creator-ab-test.md). Sequential grep-based research produces shallower, less diverse findings.
• Dual-Layer Artifacts: Every phase produces both manifest.json (machine envelope) and content.md (human-readable output). WHY: The Pipeline Architect needs machine-readable metadata to validate chain composition, while agents need readable content for reasoning. See ADR "Artifact Format: Dual-Layer Output Standard".
• Discovery Over Invention: The skill discovers subdomains through research — it does NOT hardcode or assume subdomain lists. WHY: Hardcoded lists miss domain-specific nuances and become stale. The whole point of this skill is that it adapts to any domain.
• Reuse Before Create (Rule 9): When classifying subdomains, always check if existing agents/skills cover 70%+ of the subdomain before marking it as "needs new component". WHY: Agents are expensive context. Skills are cheap. The generator biases toward binding new skills to existing agents.

Default Behaviors (ON unless disabled)

• Communication Style: Report findings without self-congratulation. Show the subdomain list and classifications directly rather than describing the process.
• Temporary File Cleanup: Phase artifacts live in /tmp/pipeline-{run-id}/. The PRODUCE phase copies final artifacts to permanent location. Intermediate files remain for debugging until the pipeline-orchestrator-engineer cleans up.
• Operator Profile Detection: Read the detected operator profile from pipeline context but do NOT gate any research steps on it. WHY: Research itself is read-only and harmless across all profiles. The profile information is passed through to the Component Manifest so downstream skills (chain-composer, scaffolder) can apply the correct safety gates.