sk-deep-research

You can automate iterative deep research using an autonomous loop that externally persists state, applying convergence checks and progressive synthesis to handle complex topics spanning multiple domains.

Externalized state in autonomous research loops persists lineage, decisions, and memory across iterations in JSONL packet artifacts, ensuring reproducible analysis and resumable sessions even with fresh-context execution.

An autonomous research loop detects convergence by evaluating ruled-out directions for negative knowledge and checking state logs, stopping iterations once the progressive synthesis output meets analytical thresholds.

Yes, the iterative research workflow provides runtime parity guidance for multiple AI runtimes including OpenCode, Claude, Codex, and Gemini to execute the research lifecycle.

You can resume an interrupted deep research session because the loop preserves memory through externalized state logs and registry artifacts, allowing the lifecycle to pick up from the last captured iteration.

An autonomous deep research loop generates canonical packet artifacts including configuration, state logs, strategy, registry, dashboard, and per-iteration notes for reproducible analysis.