a-share-research-product-workflow

name

a-share-research-product-workflow

description

Use when users need productized A-share research workflows in Codex, including watchlist daily reviews, market context reviews, daily/weekly/monthly technical analysis, explicit research versus decision_support mode routing, investment decision-support framing, strategy checks, local JSONL history, thesis review logs, backtest-style post reviews, report integrity checks, or .research artifact conventions while preserving source logs, QA, and no target-price/rating/buy-sell guardrails.

A-Share Research Product Workflow

Use this product-layer skill after financial-services-skill-router, .agents/SKILLS_INDEX.md, china-market-overlay, and a-share-equity-research-workflow when the user asks for a productized A-share research workflow rather than a one-off single-stock note.

This skill orchestrates existing skills. It does not replace the single-stock workflow, ifind-http-api, akshare, valuation, comps, thesis, or catalyst skills.

When To Use

Use for requests such as:

• A-share watchlist daily reviews, daily stock analysis, or research dashboards.
• Daily watchlist-to-decision pipelines that first screen Top10, deep-research the default Top5, and send only evidence-sufficient names into conditional decision support.
• Market context reviews before single-stock analysis.
• Daily, weekly, and monthly technical analysis in watchlists, research, and decision-support outputs.
• Explicit research mode versus decision_support mode routing.
• Investment decision-support framing without target prices, ratings, personal position sizing, or buy/sell recommendations.
• Conditional strong planning with concrete trigger, invalidation, and risk-control levels when the user asks for buy/sell-point style support, while preserving the no target-price/rating/personal-sizing guardrails.
• Vibe-Trading adapter workflows for backtest validation, Alpha Zoo screening, trade journal review, and Shadow Account summaries through trading_core.
• Strategy checks such as trend, event-driven, growth quality, valuation digestion, or thesis falsification.
• Historical report references, thesis JSONL logs, and post-review/backtest style evaluations.
• Report integrity checks for source, date, unit, QA, and guardrails.

Required Skill Order

1. financial-services-skill-router
2. .agents/SKILLS_INDEX.md
3. china-market-overlay
4. a-share-equity-research-workflow
5. This skill
6. Data and downstream skills as needed:
   • ifind-http-api
   • akshare
   • a-share-valuation-template
   • a-share-comps-best-practice
   • vertical-equity-research-thesis-tracker
   • vertical-equity-research-catalyst-calendar

Reference Files

• Prefer a matching workflow recipe in .agents/workflows/ before assembling a productized output manually.
• Use .agents/workflows/daily_a_share_decision_pipeline.json when the user wants the full watchlist -> deep research -> evidence gate -> decision support loop while preserving the individual sub-workflows.
• Read .agents/references/output-contract.md for shared cross-skill output fields.
• Read .agents/references/data-capability-registry.md before selecting iFinD, AKShare, local files, or Vibe fallback sources.
• Read .agents/references/display-profiles.md for App, CLI, JSON, or audit presentation rules.
• Read .agents/references/local-research-artifacts.md before writing .research/ artifacts.
• Read .agents/references/watchlist-management.md when the task involves creating, editing, grouping, or using local watchlists from Codex desktop conversations.
• Read references/product-workflow.md for watchlist, market context, decision-support, strategy-check, and historical-review paths.
• Read references/local-artifacts.md before writing or reading .research/ JSONL files or local Markdown reports.
• When the output will be shown directly in the Codex app conversation and trading-decision-engine is selected, use that skill's references/app-conversation-display.md for the final readable layout.
• Use the tools under tools/ when deterministic JSONL storage, report integrity checks, or post-review summaries are needed.
• Use financial-output-qa-gate before delivering or storing high-risk decision-support, backtest, alpha, journal, or valuation-adjacent outputs.

Core Rules

• Require explicit analysis_mode: research or decision_support.
• Default to research only when the user has not explicitly asked for investment decision support.
• In research, never output a decision_state.
• In decision_support, use only:
  • watch_only
  • research_candidate
  • hold_monitor
  • risk_control_review
  • avoid_or_wait
• conditional_trade_plan is allowed only in decision_support mode and must be conditional, sourced, and paired with invalidation and risk-control logic.
• Do not output target prices, buy/sell ratings, personal position sizing, return promises, unsourced consensus, or unsourced guidance.
• Technical analysis can support evidence, trigger conditions, invalidation conditions, and confidence, but it must not be the sole basis for a decision_state.
• Store proprietary or personal local history under .research/, which should remain gitignored.
• Every productized output must include source logs, data limitations, QA checks, and Not investment advice when it relates to investment decisions.

Output Requirements

Productized outputs should include the objects relevant to the task:

• analysis_mode
• source_capability_matrix
• market_context
• security_master
• source_log
• market_snapshot
• technical_analysis when market data is used for stock research or decision support
• financial_snapshot when sourced
• valuation_snapshot when sourced
• peer_set when evaluated
• thesis_tracker when thesis evidence is tracked
• decision_support only in decision_support mode
• decision_card only in decision_support mode when productized decision packaging is requested
• conditional_trade_plan only in decision_support mode when the user asks for conditional trigger/risk levels
• backtest_validation when a Vibe bridge or local validation run is sourced
• report_integrity_status
• review_history_ref when local JSONL or report files are used
• qa_status