Daily Action Plan

Daily Action Plan

Generate a daily pre-market or post-market action plan by comparing the latest weekly blog strategy with real-time market data.

Triggers

• User says "daily action plan", "DAP", "daily plan", "デイリーアクション", "寄り付きプラン", "引け後プラン"
• User says "run daily-action-plan"
• claude -p "Run daily-action-plan --timing {pre-market|post-market}"

Arguments

• --timing {pre-market|post-market} (default: auto-detect based on current time)
  • pre-market: Before 9:30 AM ET (23:30 JST winter / 22:30 JST summer)
  • post-market: After 4:00 PM ET (6:00 JST winter / 5:00 JST summer)

Workflow

Step 1: Holiday Check

python3 .claude/skills/daily-action-plan/scripts/build_plan_state.py --check-only

• CLOSED → Output a brief "Markets Closed" message with next trading day and stop.
• EARLY_CLOSE → Continue with early close note.
• OPEN → Continue normally.

Step 2: Data Fetching (only if market is open)

Fetch market data and breadth data in parallel:

python3 scripts/fetch_market_close.py --json > /tmp/dap_market.json
python3 .claude/skills/breadth-chart-analyst/scripts/fetch_breadth_csv.py --json > /tmp/dap_breadth.json

CRITICAL: These scripts must complete successfully. If either fails, report the error and stop.

Step 3: Build plan_state.json

python3 .claude/skills/daily-action-plan/scripts/build_plan_state.py \
    --timing {pre-market|post-market} \
    --market-json /tmp/dap_market.json \
    --breadth-json /tmp/dap_breadth.json \
    --output /tmp/plan_state.json

Step 4: Verify plan_state.json

python3 .claude/skills/daily-action-plan/scripts/verify_plan.py \
    --plan-state /tmp/plan_state.json \
    --market-json /tmp/dap_market.json \
    --breadth-json /tmp/dap_breadth.json

• PASS → Continue to Step 5.
• FAIL → Display errors. Investigate and fix if possible, otherwise stop.

Step 5: Generate Action Plan

Read plan_state.json and generate the action plan document in Japanese.

Reference files to read before writing:

• .claude/skills/daily-action-plan/references/monty_thresholds.md — threshold values
• .claude/skills/daily-action-plan/references/tone_guidelines.md — tone rules
• .claude/skills/daily-action-plan/assets/action_plan_template.md — output template

Writing rules:

1. Use ONLY values from plan_state.json — never invent or round numbers
2. VIX evaluation uses Monty thresholds: 17(Risk-On)/20(Caution)/23(Stress)/26(Panic)
3. Include trigger distances for each scenario
4. List today's events from plan_state.json > events > todays_events
5. Include the relevant checklist (morning for pre-market, evening for post-market)
6. Follow tone_guidelines.md strictly — no "confirmed" language, no rounding
7. Phase status from blog must be prominently displayed
8. Trigger status must use progress field from plan_state.json — never manually classify as "達成中" for weekly conditions or omit day-count for consecutive conditions

Pre-market specifics:

• Data is extended-hours/pre-market quote; previous close is used for comparison only
• Trigger progress is provisional (based on current quote, not confirmed close)
• Focus on: morning checklist, today's events, key levels to watch
• Include recommended actions based on blog scenarios

Post-market specifics:

• Use official closing data
• Evaluate trigger conditions against actual closes
• Determine which scenario is currently closest
• Include evening checklist, tomorrow's focus

Step 6: Self-Check

Verify the generated plan against plan_state.json:

1. Every number in the output must match plan_state.json exactly
2. VIX/10Y evaluations must match threshold classifications
3. Scenario probabilities must match blog values
4. Allocation percentages must match blog values

If any mismatch is found, correct and re-verify (max 3 attempts).

Step 7: Save Output

mkdir -p reports/YYYY-MM-DD

Save to: reports/YYYY-MM-DD/daily-action-plan-{pre|post}.md

IMPORTANT: Do NOT send email or call send_dap_email.py. Email notification is handled by the wrapper script (run_daily_action_plan.sh) after this skill completes. Calling it here would cause duplicate emails.

Output Format

See assets/action_plan_template.md for the full template.

Key sections:

1. Market Summary — Current prices with evaluations
2. Breadth Data — 200MA, 8MA, Uptrend Ratio with classifications
3. Scenario Distance — How far current values are from each scenario's triggers
4. Today's Events — Events for today (including daily events like geopolitical monitoring)
5. Recommended Actions — Based on blog strategy and current data
6. Checklist — Morning or evening checklist from blog

Data Flow

scripts/fetch_market_close.py --json → /tmp/dap_market.json
fetch_breadth_csv.py --json         → /tmp/dap_breadth.json
blogs/YYYY-MM-DD-weekly-strategy.md → (parsed by strategy_parser)
                                    ↓
build_plan_state.py → /tmp/plan_state.json
                    ↓
verify_plan.py → PASS/FAIL
                ↓
[Claude generates plan from plan_state.json]
                ↓
reports/YYYY-MM-DD/daily-action-plan-{pre|post}.md

Dependencies

• scripts/fetch_market_close.py — FMP API (requires FMP_API_KEY in .env)
• .claude/skills/breadth-chart-analyst/scripts/fetch_breadth_csv.py — TraderMonty CSV
• trading/layer2/tools/strategy_parser.py — Blog parser
• trading/core/holidays.py — US market holiday calendar

Error Handling