Month End

Monthly Review (global, MCP-only)

Replaces the old project-local /month-end (now archived at Journals/.claude/skills-legacy/month-end/). CLAUDE.md at /Users/latuconsinafr/Personal/Trading/Journals/CLAUDE.md remains the rulebook — this is the first skill in the set to actually exercise §13's cascade rule for real: this reads the level below (weekly rollups), not raw daily data. If a specific number looks wrong, the §13 drill-down fallback still applies — open the underlying week's data via get_daily_analysis/ get_compliance_scores rather than guessing, but that's the exception, not the default path.

If anything this session suggests a rule should change, per CLAUDE.md §20.14: add it to CLAUDE.md §21.3, do not act on it.

Step 0 — confirm the MCP server is reachable

Call get_account_state first. If it fails, stop and explain just run needs to be running.

Step 1 — resolve the month and find its weeks

Use the current YYYY-MM, unless the pilot gives one explicitly.

Call get_journal_rollups (period: "weekly", limit: 6 covers any calendar month with room to spare). From the returned rows, pick out the ones that belong to this month — the convention: a week belongs to the month its Monday falls in. State this explicitly if a week spans the boundary (e.g. starts in the prior month) rather than silently applying it; there's no single universally-correct definition for a week that spans two months, so being explicit about which convention was used matters more than the convention itself.

If a week that should exist has no row (a Monday inside this month with no matching periodKey), that week's week-end was never run. Do not backfill it — note the gap explicitly in the monthly overview, same "never silently shrink the sample" principle as week-end Step 2. State how many of the month's weeks actually have a complete rollup.

Step 2 — run the review

1. Aggregate compliance % and R-multiple trend for the month — from the weekly rollups found in Step 1, not raw daily/compliance data (that's exactly what the cascade rule exists to avoid re-reading every time). Verdict, execution, and session completeness reported separately, never merged:
   • complianceOverall/complianceVerdict/complianceExecution: average across the included weeks (each is already a rate/percentage, not additive).
   • totalR: sum across the included weeks (R is additive across time).
   • Session completeness: report as "N of M weeks had a complete rollup" (Step 1's gap check), not a percentage of daily sessions — that granularity already lives in each week's own figure.
2. Produce the weekly-timeframe monthly overview — synthesize from the included weeks' overviewMd content, not fresh chart analysis.
3. Note (do not adopt) any pattern worth flagging for the quarterly review, and confirm CLAUDE.md §21.3 is current (read it; this step doesn't require anything be added, just that it's been looked at).

Step 3 — write through MCP (the only write path)

write_journal_rollup — period: "monthly", this month's periodKey ("2026-08"). overviewMd (Step 2.2), reflectionMd (any monthly-level reflection beyond the weekly-level detail — optional, don't force content that's just a restatement of the weeks), the three compliance fields and totalR from Step 2.1.

If the call fails, stop and report it rather than treating the review as recorded.