weekly-mirror
Skill2 starsChanged 5 months agoMITView on GitHub
Brutally honest weekly performance mirror. Use every Sunday or when the user asks "how am I doing?", "weekly review", "am I making money?", "show me my track record", "should I keep trading?". Surfaces real win rate, average win/loss R-multiple, trading P&L vs boring-core P&L, and recommends continue / cool-off / suspend.
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adversarial-researchSkill
- auto-rebalanceSkill
- cash-deploymentSkill
- daily-briefingSkill
- dividend-strategySkill
- earnings-analyzerSkill
- earnings-postmortemSkill
- employer-stockSkill
- health-checkSkill
- macro-impactSkill
- momentum-scannerSkill
- optimize-allocationSkill
- pead-trackerSkill
- perf-optimizerSkill
- portfolio-healthSkill
- portfolio-reviewSkill
- position-reviewSkill
- pre-trade-checkSkill
- profile-setupSkill
- risk-assessmentSkill
- sector-rotationSkill
- stock-analysisSkill
- stock-compareSkill
- tax-loss-reviewSkill
- top-trades-todaySkill
- trade-journalSkill
- trading-deskSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 9 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Read .claude/context/STATE.md.
- Distinctive: line 18
Wait for user confirmation before continuing.
- Distinctive: line 24
After any 7-day streak of losses
- Distinctive: line 25
Before any decision to increase position sizes
- Distinctive: line 29
Before the review, load prior decisions so the mirror reflects the logged record, not memory:
- Very specific: line 47
Always pull capital from get_profile - never hardcode
- Very specific: line 132
After completing the mirror review, update .claude/context/STATE.md:
- Distinctive: line 139
NO hedging language. "Suspend discretionary" not "you might consider reducing"
- Distinctive: line 140
Always show the boring-core counterfactual (what would XEQT have done?)
- Distinctive: line 141
If 90d window has < 5 closed trades, mark stats as "insufficient sample" but still show 7d/30d
- Distinctive: line 143
If user disputes verdict, re-state thresholds.
- Distinctive: line 155
Do NOT recommend specific replacement trades in this skill.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Weekly Mirror (Cold Performance Review)
Goal
One scroll-length, no-sugarcoat performance review of the user's discretionary trading vs their boring-core (index / ETF) holdings. Tells the user the truth even when uncomfortable. Recommends one of: continue, cool-off 30 days, suspend discretionary.
The point: most retail traders never measure their real win rate. They remember winners, forget losers, and feel like they're doing fine until the account is down. This skill makes the math impossible to ignore.
State check (BEFORE any tool calls)
Read .claude/context/STATE.md. If last_mirror_date is within the last 6 days, output:
"Weekly mirror ran on
last_mirror_date- only N days ago. Run again? (yes to proceed)" Wait for user confirmation before continuing. If user says yes or forced, proceed normally.
When to invoke
- User asks "how am I doing?", "am I making money?"
- Sunday evening review (can be scheduled via /loop)
- After any 7-day streak of losses
- Before any decision to increase position sizes
Stage 0 - Decision Memory (load FIRST)
Before the review, load prior decisions so the mirror reflects the logged record, not memory:
mcp__aifolimizer__get_cross_ticker_lessonswithmax_lessons=3- portfolio-level win/loss patterns- For any name you single out, load
mcp__aifolimizer__get_ticker_decision_history(ticker=…, max_decisions=5) andmcp__aifolimizer__get_ticker_reflection(symbol=…, n=3).
How to run
Step 1 - Pull state (parallel):
mcp__aifolimizer__get_profile- total NAV per account, cash balancesmcp__aifolimizer__get_portfolio- current holdings + day/total returnsmcp__aifolimizer__score_recommendations- mark-to-market all open recs frompre-trade-checkand other skills, mark stops/targets hitmcp__aifolimizer__get_live_track_recordwithwindows_days=[7,30,90](single call returns all three windows) - win rate + P&L per windowmcp__aifolimizer__get_alpha_attributionwithlookback_days=90- am I beating the index? (benchmarks are fixed internally: SPY / XEQT / TSX / QQQ - the tool takes nobenchmarksarg)mcp__aifolimizer__snapshot_portfolio_equity- append today's NAV to history (idempotent per day)mcp__aifolimizer__get_cross_ticker_lessonswithmax_lessons=5- recurring patterns from prior stop-outsmcp__aifolimizer__get_personal_context- account waterfall + contribution room so DCA/verdict actions are tailored. Ifpresent=false, skip personalization gracefully.
Investor profile
- Always pull capital from
get_profile- never hardcode - Boring-core tickers default:
["XEQT.TO", "VFV.TO", "VTI", "QQQ", "XIC.TO", "ZSP.TO"]. Treat as passive bucket. - Trading tickers = everything else in portfolio. Treat as discretionary bucket.
Output structure
Lead with the scoreboard. Save context for the verdict.
1. The Scoreboard (top of brief)
WEEKLY MIRROR · <DATE> · <ACCOUNT or AGGREGATE>
================================================
Total NAV: $X,XXX (Δ this week: ±A.A% / ±$BBB)
Total deposits YTD: $X,XXX
Return vs deposits: ±C.C% ← THIS IS THE NUMBER THAT MATTERS
7-day P&L breakdown:
Boring-core (ETFs/index): ±$D (B.B%)
Discretionary trades: ±$E (E.E%)
Difference: ±$F ← are trades helping or hurting?
vs SPY this week: ±G.G%
vs XEQT.TO this week: ±H.H%2. Trade Track Record (table)
| Window | Closed Trades | Win Rate | Avg Win | Avg Loss | R-multiple | Net P&L |
| 7d | N | X% | $X | $X | X.XR | ±$X |
| 30d | N | X% | $X | $X | X.XR | ±$X |
| 90d | N | X% | $X | $X | X.XR | ±$X |
Interpret R-multiple:
- R ≥ 2.0 → entries+exits are working (rare for retail)
- 1.0 ≤ R < 2.0 → break-even after costs, need higher win rate
- R < 1.0 → losers larger than winners - classic discipline failure
- Win rate < 40% + R < 1.5 → STOP
3. Open positions snapshot (from score_recommendations)
- Open recs count, total at-risk capital, total floating P&L
- Stops hit but not exited (action: close now)
- Targets hit but not trimmed (action: trim 50% now)
- Time-decay positions (>30 days open without resolution - review thesis)
4. Recurring patterns (from cross-ticker lessons)
List up to 3 repeat-mistake patterns the user is making. Examples:
- "5 of last 8 stop-outs were single-stock semiconductors entered after >10% week"
- "All 4 losing trades in last 30d had crowding_score >70 at entry"
- "Average hold time on winners: 2.1 days. On losers: 14.8 days. Holding losers too long."
This is the most valuable section - repeated mistakes are the only fixable thing.
5. Boring-core check
- Total in boring-core (XEQT/VFV/etc): $X (Y% of NAV)
- Auto-DCA active? (check if user has recurring buys)
- Boring-core YTD return: ±%
- Discretionary YTD return: ±%
- Honest math: would moving 100% to XEQT 1 year ago have outperformed? Compute using boring-core return and apply to total NAV.
6. Verdict (REQUIRED - no soft language)
One of:
🟢 CONTINUE Criteria: 30d win rate ≥ 50% AND R-multiple ≥ 1.5 AND discretionary beating boring-core by ≥3% over 90d. Output: "Discipline is working. Maintain position sizes."
🟡 COOL-OFF 30 DAYS Criteria: 30d win rate 40-50%, R-multiple 1.0-1.5, OR discretionary underperforming boring-core by 0-5% over 90d. Output: "Reduce position sizes by 50% for next 30 days. Take half the trades. Journal every entry."
🔴 SUSPEND DISCRETIONARY
Criteria: 30d win rate < 40%, OR R-multiple < 1.0, OR discretionary underperforming boring-core by ≥5% over 90d.
Output: "Stop discretionary trading for 30 days. Allocate next 4 weeks of contributions to your core index ETF (DCA only). Revisit after 30d." If the user has no confirmed core ETF yet (no saved preference / none held), do NOT name one - tell them to run auto-rebalance first to pick a core sleeve on merit, then DCA into that.
7. Next actions (≤ 3 bullets)
- Specific calls to action with $ amounts
- Example: "Sell PLTR (stop hit 4d ago, still holding): -$340 realized > continued risk"
- Example: "Move $2,000 settled cash → XEQT.TO biweekly DCA"
- Example: "Skip next 5 swing-trade temptations. Journal them but do not enter."
After output - write STATE.md
After completing the mirror review, update .claude/context/STATE.md:
last_mirror_date: today's date (YYYY-MM-DD)open_recs: count of open recommendations fromscore_recommendations
Rules
- ≤ 500 words total
- NO hedging language. "Suspend discretionary" not "you might consider reducing"
- Always show the boring-core counterfactual (what would XEQT have done?)
- If 90d window has < 5 closed trades, mark stats as "insufficient sample" but still show 7d/30d
- Verdict is determined by math thresholds above, not feel. State the math
- If user disputes verdict, re-state thresholds. Do not flip on argument
- Currency: report in CAD (account base), mark USD positions explicitly
Gotchas
score_recommendationsonly sees recs that were logged vialog_recommendation. If user trades without invokingpre-trade-check, this skill is blind to those trades. Recommend: enforcepre-trade-checkon every entryget_live_track_recordcached per session - for fresh stats, callscore_recommendationsfirst to mark-to-market open positionsget_alpha_attributionneedssnapshot_portfolio_equityhistory. New users with <30 days of history will see "insufficient history" - note in output- Wealthsimple Managed account returns are not user-controlled - exclude from discretionary attribution. Only attribute self-directed accounts (TFSA-self-directed, RRSP-self-directed, Non-Reg-self-directed)
- Day-trade P&L: if user has intraday closes,
score_recommendationsmay not see them because rec was opened+closed same session. Document this as known blind spot until intraday-rec logging exists - Boring-core tickers list is configurable - if user holds different ETFs (e.g. VOO, SCHD), update the default list in Step "Investor profile"
- Crypto P&L (CADC, BTC, etc.) - treat as separate bucket, not boring-core, not discretionary. Volatility skews trade stats
- Do NOT recommend specific replacement trades in this skill. This is a mirror, not an advisor. Trade picking belongs in cash-deployment or stock-analysis
- Verdict thresholds calibrated for typical retail. For users with prop-trading background, ask if they want tighter thresholds
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.

