Inspect prior analysis records and artifacts. Use when user wants to browse, compare, or track how saved views changed over time rather than just reading one stored file.
Run a deeper one-symbol A-share research workflow. Use when user wants a research memo that goes beyond the default brief and pressure-tests thesis, catalysts, risks, and invalidation using persisted artifacts.
Evaluate a saved analysis run against later market data. Use when user wants retrospective review grounded in a saved state or a deterministic strategy simulation with explicit assumptions.
Run or inspect data synchronization workflows. Use when user wants the stage-2 data-acquisition control surface: freshness checks, provider health, source comparison, stale-artifact diagnosis, and explicit next actions.
Build a batch decision dashboard for multiple symbols. Use when user wants the stage-6 watchlist action surface: action distribution, focus list, and portfolio-assumption-aware triage before deeper single-name review.
Produce action guidance with sizing and risk framing. Use when user wants a conditional buy, hold, reduce, avoid, or watch decision tied to explicit account assumptions.
Push one explicit Feishu notification card. Use when a workflow needs the stage-7 communication step: outbound delivery of a status, checkpoint, or recap artifact under explicit fail-open semantics.
Autonomous error recovery and workflow continuation. Use when any skill fails, data is missing, or workflow is blocked. Agent analyzes root cause, attempts fixes, and continues workflow without user intervention. Designed for non-technical users who should not see internal errors.