paper-trading-post-market-review
Skill0 starsChanged 2 months agoApache-2.0View on GitHub
Use when a workshop agent is reviewing an A-share paper-trading day after market close, especially to reconcile account results, holdings, orders, plan-versus-actual behavior, mistakes, missed triggers, data-source quality, reusable trading rules, and next-day preparation.
In this repository20 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-browserSkill
- ai-knowledge-creatorSkill
- docxSkill
- frontend-designSkill
- harness-quality-controlSkill
- openzhiyu-apiSkill
- openzhiyu-feature-guideSkill
- openzhiyu-memorySkill
- paper-trading-intraday-checkSkill
- paper-trading-pre-market-planSkill
- paper-trading-trend-followingSkill
- pdfSkill
- pptxSkill
- single-etf-position-advisorSkill
- single-etf-timing-advisorSkill
- ui-ux-pro-maxSkill
- watchlist-selection-controlSkill
- weatherSkill
- xlsxSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
27 rules. About 0 are common, 20 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈20
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Use this for the 收盘后模拟复盘 loop or any similar A-share paper-trading after-close review.
- Distinctive: line 19
Do not rewrite history to fit the close.
- Distinctive: line 23
Call tools in this order when available:
- Distinctive: line 26
Use the authoritative local date and weekday for report titles.
- Very specific: line 27
Do not infer weekdays from memory.
- Distinctive: line 30
Read final account, cash, positions, open orders, fills, realized PnL, unrealized PnL, and total asset.
- Distinctive: line 34
Read close-state watchlist quotes.
- Distinctive: line 38
Read today's active and historical plans before judging process quality.
- Distinctive: line 42
Read today's pre-market plan, intraday memories, active rules, prior mistakes, and position-specific triggers.
- Distinctive: line 43
If a memory drove an action or rule update, verify evidence when possible.
- Distinctive: line 46
Read today's same-workshop events when available, especially pre-market and intraday loop events.
- Distinctive: line 50
Use for held positions, executed orders, and symbols that triggered action/no-action decisions.
- Distinctive: line 54
Use for held positions and key watchlist symbols.
- Distinctive: line 61
Use for held positions, executed orders, abnormal moves, and material next-day risks.
- Distinctive: line 81
If numbers conflict across sources, stop and log the conflict before drawing conclusions.
- Distinctive: line 90
If it acted, was the action inside the rule?
- Distinctive: line 91
If it did not act, was the no-action reason valid?
- Distinctive: line 94
Do not judge decisions solely by closing PnL.
- Distinctive: line 150
When writing memory, include:
- Distinctive: line 160
Prefer confidence 55-70 for single-day lessons and 75+ only for repeated patterns.
- Distinctive: line 172
Avoid predicting next-day direction.
- Very specific: line 176
Before judging whether a trade plan, stop, invalidation, breakout, cash, position-size, profit-protection, or risk/reward rule triggered or …
- Very specific: line 178
Use quantRuleEvaluate to reconcile due plans before calling quantTradePlanReview.
- Very specific: line 179
Every reviewed plan that is marked executed, blocked, skipped, missed_trigger, invalidated, or superseded because of a number must include …
- Distinctive: line 180
If the rule cannot be checked due to missing or degraded data, preserve that as the blocker and do not invent the numeric outcome.
- Very specific: line 182
Before finishing, call quantTradePlanReview for today's plans and call quantTradePlanUpsert for concrete next-day plans.
- Very specific: line 186
Before finishing, call workshopLogEvent with these exact title prefixes:
≈ “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.
Files
Paper Trading Post-Market Review
This skill turns the post-market run into feedback and learning. The agent should evaluate process quality, not merely explain the day's PnL.
Use this for the 收盘后模拟复盘 loop or any similar A-share paper-trading after-close review.
Core Philosophy
- Separate outcome from process.
- A profitable day can still contain a bad decision.
- A losing day can still contain correct risk control.
- Convert repeatable lessons into workshop memory.
- Weaken or correct memories when new evidence contradicts them.
- Do not rewrite history to fit the close.
Required Observation Order
Call tools in this order when available:
- Current execution time context or
time- Use the authoritative local date and weekday for report titles.
- Do not infer weekdays from memory. For example,
2026-08-05must be rendered as2026-08-05(周三).
quantPaperGetAccount- Read final account, cash, positions, open orders, fills, realized PnL, unrealized PnL, and total asset.
- Reconcile whether any order executed, failed, or remained open.
quantPaperGetWatchlist- Read close-state watchlist quotes.
- Compare held positions with watchlist leaders and laggards.
quantTradePlanList- Read today's active and historical plans before judging process quality.
- Reconcile each due plan as executed, partially executed, blocked, skipped, superseded, or still pending with reason.
workshopReadMemoryorworkshopSearchMemory- Read today's pre-market plan, intraday memories, active rules, prior mistakes, and position-specific triggers.
- If a memory drove an action or rule update, verify evidence when possible.
workshopReadLinkedWorkshopEvents- Read today's same-workshop events when available, especially pre-market and intraday loop events.
- Reconcile plan versus actual behavior.
aStockTrend- Use for held positions, executed orders, and symbols that triggered action/no-action decisions.
- Review whether the entry, hold, reduce, or sell decision matched trend state, trendScore, MA/ATR stop, and extension.
aStockSignals- Use for held positions and key watchlist symbols.
- Record source limitations and conflicts explicitly.
aStockMarketMood- Summarize market breadth, theme leadership, and whether conditions validated or invalidated the day's plan.
aStockNewsAndFilings- Use for held positions, executed orders, abnormal moves, and material next-day risks.
- News is a review input, not a retroactive justification.
Review Framework
1. Account Reconciliation
Report:
- total asset,
- cash,
- market value,
- realized PnL,
- unrealized PnL,
- position count,
- open orders,
- filled orders,
- day change,
- exposure by symbol and theme.
If numbers conflict across sources, stop and log the conflict before drawing conclusions.
2. Plan Versus Actual
For each plan or trigger from the pre-market/intraday events:
- Was it observed?
- Did it fire?
- Did the agent act?
- If it acted, was the action inside the rule?
- If it did not act, was the no-action reason valid?
- Did data quality block a decision?
Do not judge decisions solely by closing PnL.
3. Signal Evaluation
Classify each held position:
validated: price, relative strength, fund flow, and news supported the plan.trend_validated: trend state and score still support the holding thesis.trend_warning: price remains tradable but stop distance, extension, or MA structure is deteriorating.warning: one major signal diverged.invalidated: predefined stop, thesis, or risk rule failed.uncertain: source conflict or insufficient data prevents judgment.
Classify watchlist context:
- leaders,
- constructive pullbacks,
- laggards,
- overheated names,
- names to ignore tomorrow.
This skill does not maintain the watchlist. If a watchlist action is needed, record a memory for the watchlist-selection workshop rather than calling watchlist proposal tools.
4. Mistake Taxonomy
Use this taxonomy:
chasing: acted after risk/reward was gone.premature_exit: exited before invalidation without evidence.ignored_stop: failed to obey a hard risk rule.data_quality: acted on stale, degraded, or contradictory data.news_overweight: let news dominate price/position evidence.overfitting: made a new rule from one weak example.missed_trigger: valid trigger fired but no action occurred.trend_misread: trend state, extension, or MA/ATR stop was misread.good_no_action: did nothing despite market emotion.good_risk_control: controlled downside even if PnL later improved.
5. Memory Update Rules
Use workshopWriteMemory for:
- a repeated pattern with evidence,
- a specific rule that should constrain future actions,
- a source-quality warning,
- an active position trigger,
- a mistake that should be checked next time,
- a contradicted old rule that needs weakening.
Do not write:
- generic market summary,
- one-off price moves without lesson,
- celebratory PnL notes,
- speculative claims without evidence.
When writing memory, include:
- date,
- symbol or scope,
- evidence,
- rule,
- invalidation,
- next-use condition,
- confidence.
Prefer confidence 55-70 for single-day lessons and 75+ only for repeated patterns.
Next-Day Preparation
End with a next-day control plan:
- account risk to check first,
- position-specific triggers,
- data sources that need verification,
- no-trade conditions,
- maximum action scope for the next loop.
Avoid predicting next-day direction.
Numeric Rule Discipline
Before judging whether a trade plan, stop, invalidation, breakout, cash, position-size, profit-protection, or risk/reward rule triggered or failed, batch the comparisons and call quantRuleEvaluate.
- Use
quantRuleEvaluateto reconcile due plans before callingquantTradePlanReview. - Every reviewed plan that is marked
executed,blocked,skipped,missed_trigger,invalidated, orsupersededbecause of a number must includeruleEvidencewithtool: "quantRuleEvaluate",ruleId,actual,operator, threshold or range,triggered, andcomparisonText. - If the rule cannot be checked due to missing or degraded data, preserve that as the blocker and do not invent the numeric outcome.
Before finishing, call quantTradePlanReview for today's plans and call quantTradePlanUpsert for concrete next-day plans. If a plan was not completed, preserve the exact blockerReason rather than replacing it with a generic summary.
Required Visible Output
Before finishing, call workshopLogEvent with these exact title prefixes:
marketScanSummary: ...trendStateReview: ...riskAssessment: ...actionTaken: ...
Also use one of these prefixes when applicable:
processReview: ...memoryUpdate: ...nextDayPlan: ...
Final JSON Contract
The final <structured-output> JSON must include:
{
"marketScanSummary": "account, positions, watchlist, market breadth and material news summary",
"riskAssessment": "process risk, position risk, data quality and rule conflicts",
"actionTaken": "orders/fills reviewed, memories written, no operation, or blocked reason",
"tradePlanLedger": "plan-versus-actual reconciliation and next plans written",
"postMarketReview": {
"processGrade": "A|B|C|D|F",
"validatedRules": [],
"weakenedRules": [],
"newMemories": [],
"nextDayPlan": []
}
}Reference Basis
For deeper rationale, read references/methodology.md. Use it when deciding whether a lesson deserves memory or whether a rule is overfit.
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
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.

