Index of every page on this site: llms.txt
All skills
tradermonty avatar

/weekly-performance-digest

@f95e770
Run it yourself

Generate a weekly performance summary from closed trader-memory-core theses — win rate, expectancy, profit factor, R-multiple, MAE/MFE, and win/loss pattern analysis by source skill, exit reason, thesis type, sector, and mechanism. No API required; pure local calculation.

Use this Skill: https://skilld.dev/gh/tradermonty/claude-trading-skills/weekly-performance-digest

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈75 tokens for metadata: the name and description. ≈1.5k when used: this file. ≈949 more on demand in 1 file.

Description uses 3.8% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈300 characters, or 3.8%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Weekly Performance Digest

Overview

Weekly Performance Digest aggregates the trades you closed during a week into a single performance report. It reads CLOSED theses tracked by trader-memory-core (state/theses/th_*.yaml), computes headline metrics (win rate, expectancy, profit factor, R-multiple, MAE/MFE), breaks results down across several pattern dimensions (source skill, exit reason, thesis type, sector, mechanism tag, screening grade), and surfaces the week's biggest winners, losers, and lessons. Output is a JSON record plus a human-readable Markdown report. Pure calculation — no API key required.

R-multiples require positive risk from the recorded stop. Gap fills at/below the stop and stops raised to/above entry have no R-multiple; these trades remain in P&L and win/loss statistics but are excluded from R averages.

When to Use

  • At the end of a trading week to review aggregate realized performance
  • To measure win rate and expectancy across all closed positions
  • To see which source skills, exit reasons, sectors, or mechanisms drove wins vs losses
  • To feed a month-end review (combine four weekly digests) or a postmortem
  • For a quick "what worked / what didn't" snapshot grounded in real closed trades

When Not to Use

  • For a single-trade deep review — use trade-performance-coach
  • For signal-level true/false-positive classification — use signal-postmortem
  • For buy/sell recommendations or position sizing — this skill is descriptive only

Prerequisites

  • Python 3.9+ with PyYAML (already a repo dependency)
  • A trader-memory-core state directory of thesis YAML files (state/theses/)
  • No API key required

Workflow

Step 1: Run the digest for a week

python3 skills/weekly-performance-digest/scripts/generate_weekly_digest.py \
  --state-dir state/theses \
  --from-date 2026-06-13 --to-date 2026-06-20 \
  --output-dir reports/ -v

Defaults: --state-dir state/theses, --from-date = 7 days before --to-date, --to-date = today, --output-dir reports/. With no date flags it digests the trailing 7 days.

Step 2: Read the report

The run writes reports/weekly_digest_<to-date>.json and reports/weekly_digest_<to-date>.md. Review the Markdown for the executive summary, metrics table, pattern breakdowns, and top winners/losers; consume the JSON downstream.

Step 3 (optional): Feed downstream

Combine several weekly JSON digests for a monthly review, or pass the JSON to a postmortem/coach step. The skill is descriptive — act on its findings via your normal review process.

How It Works

  • Trade selection. A trade counts in a week if its exit.actual_date falls in [from-date, to-date] and status == CLOSED.
  • Win/loss. outcome.pnl_dollars > 0 is a winner, < 0 a loser, == 0 breakeven; win_rate = winners / total_trades.
  • R-multiple. pnl_dollars / ((entry.actual_price − exit.stop_loss) × position.shares). Calculated only when the denominator is positive. A gap fill at/below the recorded stop or a stop raised to/above entry makes R unavailable (null in JSON, n/a in Markdown), so it is excluded from R averages. The trade still counts in P&L, win/loss statistics, and top wins/losses. The recorded stop is read from exit.stop_loss and may differ from the initial planned stop.
  • Double-counting safeguard. A CLOSED thesis's outcome.pnl_dollars is the cumulative realized P&L across all trims plus the final leg. Headline metrics use that cumulative value over CLOSED theses only. The separate partial_trims block scans status_history[] of PARTIALLY_CLOSED theses only (still open) and is reported for information — it is never added into the headline totals/win-rate. A position trimmed in week 1 then closed in week 2 therefore shows as a partial trim in week 1 and inside week 2's CLOSED headline; that is intended, not a duplicate.

Output Format

JSON (weekly_digest_<to-date>.json)

{
  "schema_version": "1.0",
  "report_type": "weekly_performance_digest",
  "period": {"from": "2026-06-13", "to": "2026-06-20"},
  "generated_at": "2026-06-20T21:39:07Z",
  "summary": {
    "total_trades": 2, "winners": 1, "losers": 1, "breakeven": 0,
    "win_rate": 0.5, "expectancy": 25.0, "profit_factor": 2.0,
    "total_realized_pnl": 50.0, "total_realized_pnl_pct": 4.17
  },
  "metrics": {
    "avg_winner": 100.0, "avg_loser": -50.0,
    "largest_winner": 100.0, "largest_loser": -50.0,
    "avg_holding_days_winners": 9.0, "avg_holding_days_losers": 6.0,
    "r_multiple_avg": 0.25, "r_multiple_stdev": 1.06,
    "avg_mae_pct": -3.75, "avg_mfe_pct": 4.5
  },
  "pattern_analysis": {
    "by_source_skill": {"...": {"wins": 1, "losses": 0, "total": 1, "win_rate": 1.0}},
    "by_exit_reason": {}, "by_thesis_type": {}, "by_sector": {},
    "by_mechanism_tag": {}, "by_screening_grade": {}
  },
  "partial_trims": {"count": 0, "total_realized_pnl": 0.0, "trims": []},
  "lessons": {"top_wins": [], "top_losses": [], "process_improvements": []}
}

Markdown (weekly_digest_<to-date>.md)

Sections: # Weekly Performance Digest, ## Executive Summary, ## Performance Metrics, ## Pattern Analysis, ## Lessons Learned (### Top Winners / ### Top Losers / ### Process Improvements).

An empty week still produces a valid report with zeroed metrics (exit code 0).

Resources

  • scripts/generate_weekly_digest.py — digest generator (JSON + Markdown)
  • references/weekly-digest-metrics.md — metric formulas and interpretation

Key Principles

  1. Closed trades only for headline numbers — cumulative outcome.*, keyed on exit date.
  2. No double-counting — partial trims are informational and excluded from totals.
  3. Pattern attribution — every win/loss is attributed across multiple dimensions.
  4. Descriptive, not prescriptive — the digest reports; you decide.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts3d3 checks · Risk SAFE
  • Gen Agent Trust Hub3d

    The skill is a local utility that aggregates trading performance data from YAML files into Markdown and JSON reports. It operates entirely on the local file system and does not require network access or external API keys. A minor vulnerability surface for indirect prompt injection was identified as the skill processes external data into reports without specific sanitization, although no active malicious behavior was found.

  • Socket3d

    No alerts

  • Snyk3d

    Risk: LOW · No issues

Signed by skilld at f95e770. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week

README badge

README badge for tradermonty/claude-trading-skills/weekly-performance-digest