agentleFS
JayuanlabJayuanlab/zhiyu/skills/paper-trading-pre-market-plan/SKILL.md

paper-trading-pre-market-plan

Skill0 starsChanged 2 months agoApache-2.0View on GitHub

Use when a workshop agent is preparing an A-share paper-trading pre-market plan, especially before market open, before a simulated trading day, or when it must define observation priorities, no-trade conditions, risk boundaries, and if-then triggers for the paper account.

The file

Show its rules and text

We split this skill into the rules it gives an agent

27 rules. About 0 are common, 18 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈18

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this for the 开盘前观察计划 loop or any similar A-share paper-trading pre-open planning task.

    line 10
  2. Distinctive:

    Treat the paper account as a controlled experiment.

    line 14
  3. Distinctive:

    Do not predict the day. Prepare conditional responses.

    line 15
  4. Distinctive:

    Prefer no operation unless current account state, market regime, watchlist evidence, risk/reward, and boundary policy all agree.

    line 16
  5. Distinctive:

    Never let news, one quote, or a single fund-flow number become an action by itself.

    line 17
  6. Distinctive:

    Use yesterday's review and active workshop memory as the state estimate.

    line 18
  7. Distinctive:

    Call tools in this order when available and within budget:

    line 23
  8. Distinctive:

    Read cash, total asset, realized PnL, positions, available quantity, open orders, fills, and trading enabled state.

    line 26
  9. Distinctive:

    Stop if account data is missing, stale, or internally inconsistent.

    line 27
  10. Distinctive:

    Read current watchlist and quote snapshot.

    line 30
  11. Distinctive:

    Read active and recent trade plans before creating new triggers.

    line 34
  12. Distinctive:

    Read active trading rules, yesterday's plan, known mistakes, position-specific triggers, and source limitations.

    line 38
  13. Distinctive:

    Use breadth, limit-up/limit-down, hot themes, and large-style direction.

    line 43
  14. Distinctive:

    Use phase, trendScore, MA20/MA60 structure, ATR stop, extension, and invalidation as the primary method layer.

    line 51
  15. Distinctive:

    Use for holdings and high-priority watchlist names.

    line 54
  16. Distinctive:

    Treat fund-flow degradation or source conflicts as a risk flag, not as proof.

    line 55
  17. Distinctive:

    Use for held positions, open-order candidates, abnormal moves, or major planned decisions.

    line 58
  18. Very specific:

    Before concluding that any price, stop, invalidation, breakout, cash, position-size, profit-protection, or risk/reward rule has triggered …

    line 63
  19. Very specific:

    Use stable rule ids such as 000977_buy_invalid_line, 159278_profit_stop, or cash_position_cap.

    line 66
  20. Very specific:

    Every actionable trigger in preMarketPlan.ifThenTriggers[] must carry ruleEvidence with tool: "quantRuleEvaluate", ruleId, actual, …

    line 67
  21. Very specific:

    If quantRuleEvaluate evidence is missing, write the plan as pending_tool_check and do not claim that a trigger, stop, or invalidation fired.

    line 68
  22. Distinctive:

    Do not create a watchlist change proposal.

    line 100
  23. Very specific:

    Default action is no operation.

    line 104
  24. Very specific:

    Before finishing, call workshopLogEvent with these exact title prefixes:

    line 130
  25. Very specific:

    Use workshopWriteMemory only for reusable state:

    line 137
  26. Distinctive:

    Do not write routine summaries as memory unless they contain a reusable rule or a next-run trigger.

    line 145
  27. Very specific:

    If the run creates, revises, or carries forward a concrete plan, call quantTradePlanUpsert before finishing.

    line 147

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Paper Trading Pre-Market Plan

This skill turns the pre-market run into a control-plan pass, not a prediction pass. The agent should define what it will observe, what would invalidate action, and what must remain untouched.

Use this for the 开盘前观察计划 loop or any similar A-share paper-trading pre-open planning task.

Core Philosophy

  • Treat the paper account as a controlled experiment.
  • Do not predict the day. Prepare conditional responses.
  • Prefer no operation unless current account state, market regime, watchlist evidence, risk/reward, and boundary policy all agree.
  • Never let news, one quote, or a single fund-flow number become an action by itself.
  • Use yesterday's review and active workshop memory as the state estimate; verify high-impact memories with current data before relying on them.
  • Respect A-share constraints: T+1, price-limit regimes, lunch break, closing auction, lot-size constraints, and delayed correction after opening mistakes.

Required Observation Order

Call tools in this order when available and within budget:

  1. quantPaperGetAccount
    • Read cash, total asset, realized PnL, positions, available quantity, open orders, fills, and trading enabled state.
    • Stop if account data is missing, stale, or internally inconsistent.
  2. quantPaperGetWatchlist
    • Read current watchlist and quote snapshot.
    • Stay inside current watchlist. This skill does not discover or propose watchlist changes.
  3. quantTradePlanList
    • Read active and recent trade plans before creating new triggers.
    • Review whether prior plans are still valid, superseded, blocked, or due today.
  4. workshopReadMemory or workshopSearchMemory
    • Read active trading rules, yesterday's plan, known mistakes, position-specific triggers, and source limitations.
    • For high-impact memories, use workshopGetMemoryEvidence if uncertainty could change an action.
  5. aStockMarketMood
    • Classify the market into one of: red, yellow, green, or unknown.
    • Use breadth, limit-up/limit-down, hot themes, and large-style direction.
  6. aStockQuote
    • Refresh holdings first, then watchlist focus names.
    • Current price data beats old workshop events.
  7. aStockTrend
    • Required for held positions, open-order candidates, and focus names that may become action candidates.
    • Use phase, trendScore, MA20/MA60 structure, ATR stop, extension, and invalidation as the primary method layer.
  8. aStockSignals
    • Use for holdings and high-priority watchlist names.
    • Treat fund-flow degradation or source conflicts as a risk flag, not as proof.
  9. aStockNewsAndFilings
    • Use for held positions, open-order candidates, abnormal moves, or major planned decisions.
    • News is supporting evidence only.

Numeric Rule Discipline

Before concluding that any price, stop, invalidation, breakout, cash, position-size, profit-protection, or risk/reward rule has triggered or failed, batch the comparisons and call quantRuleEvaluate.

  • One model response may contain many numeric checks. Put them in one quantRuleEvaluate.rules[] call when possible instead of doing mental arithmetic.
  • Use stable rule ids such as 000977_buy_invalid_line, 159278_profit_stop, or cash_position_cap.
  • Every actionable trigger in preMarketPlan.ifThenTriggers[] must carry ruleEvidence with tool: "quantRuleEvaluate", ruleId, actual, operator, threshold or range, triggered, and comparisonText.
  • If quantRuleEvaluate evidence is missing, write the plan as pending_tool_check and do not claim that a trigger, stop, or invalidation fired.

Market State Model

Assign exactly one state:

red

  • Broad market is in retreat, limit-down risk is elevated, hot themes are collapsing, data sources are unreliable, or account risk is already above boundary.
  • Output: no new buy. Only plan defensive checks.

yellow

  • Market is mixed, rotating, overextended, data is partially degraded, or signals conflict.
  • Output: observe. New buys require unusually clean evidence and small size.

green

  • Breadth, leading themes, liquidity, and watchlist behavior are aligned.
  • Output: allow conditional triggers, but still require risk/reward and position rules.

unknown

  • Data freshness or source quality is insufficient.
  • Output: no operation; log the data gap.

Watchlist Classification

Classify each relevant symbol:

  • leader: stronger than market and sector, liquidity acceptable, not extended far beyond a valid base or plan.
  • constructive: improving but still waiting for confirmation.
  • laggard: underperforms its sector/market or has weak relative strength.
  • risk: has source conflict, negative event, stop-risk, liquidity issue, or gap risk.
  • ignore_today: not relevant to today's control objective.

Do not create a watchlist change proposal. That belongs to the watchlist-selection workshop.

Pre-Market Action Rules

Default action is no operation.

Allow a planned simulated buy only if all gates pass:

  • Account gate: cash, available quantity, open orders, and trading enabled state are known.
  • Market gate: market state is not red or unknown.
  • Symbol gate: target is in current watchlist.
  • Evidence gate: price/volume, relative strength, fund flow, and news context do not conflict materially.
  • Trend gate: trendScore is normally at least 65, trend state is strong_uptrend, early_breakout, or constructive_pullback, and the stop/invalidation line is explicit.
  • Risk gate: planned stop is explicit and risk/reward is at least 1:2.
  • Position gate: position size will not concentrate the account or violate existing workshop memory.
  • Execution gate: every planned buy must name a plannedPrice and an execution tolerance. Default tolerance is plannedPrice + 2%; use a limit order and explain why it is not chasing the open.

Block action when:

  • Data is stale, missing, or contradictory.
  • The plan depends on news that is unverified, promotional, old, or not material.
  • The symbol is outside the watchlist.
  • The action was invented in this run without prior setup.
  • Opening price has moved enough that the original risk/reward is gone.
  • Current executable buy price is more than 2% above the planned entry price and no fresh risk/reward plan has replaced it.
  • Trend state is range, downtrend, broken, or extended without a fresh breakout thesis.
  • The model is trying to "make something happen" because a loop ran.

Required Visible Output

Before finishing, call workshopLogEvent with these exact title prefixes:

  • marketScanSummary: ...
  • trendStateReview: ...
  • riskAssessment: ...
  • actionTaken: ...

Use workshopWriteMemory only for reusable state:

  • a concrete trigger plan that future runs must check,
  • a source-quality limitation,
  • a trading-rule correction,
  • a mistake or boundary lesson,
  • a position-specific invalidation rule.

Do not write routine summaries as memory unless they contain a reusable rule or a next-run trigger.

If the run creates, revises, or carries forward a concrete plan, call quantTradePlanUpsert before finishing. Each plan must include planDate, code, action, triggerCondition, invalidation, rationale, and dueAt when known.

Final JSON Contract

The final <structured-output> JSON must include:

{
  "marketScanSummary": "market, account, holdings, watchlist and evidence summary",
  "riskAssessment": "risk state, data quality, invalidation conditions, and why action is or is not allowed",
  "actionTaken": "simulated order submitted, no operation, or blocked with reason",
  "tradePlanLedger": "prior plans reviewed and new plans written with quantTradePlanUpsert",
  "preMarketPlan": {
    "marketState": "red|yellow|green|unknown",
    "focusSymbols": [],
    "noTradeConditions": [],
    "ifThenTriggers": []
  }
}

Reference Basis

For deeper rationale, read references/methodology.md. Use it only when the task needs method explanation or rule revision.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.