Your First Week

A seven-day path from installation to a repeatable market check, first journal entry, and first weekly review.

Table of Contents

Before You Start

This guide needs a Claude plan that supports Skills, Python 3.9+, git, uv, and internet access to public CSV files. It needs no paid market-data API and no FMP, FINVIZ Elite, or broker credentials.

“No paid API” does not mean “offline.” The two market analyzers download public CSV files. The journal and weekly review stay local. Nothing in this guide places an order or turns a report into a buy/sell signal.

The copy-and-paste commands below are for Claude Code or a terminal at the repository root. Web App users can upload the corresponding .skill files from skill-packages/: trading-skills-navigator, market-breadth-analyzer, uptrend-analyzer, exposure-coach, trader-memory-core, and weekly-performance-digest.

Day 1 — Install a Reproducible Environment

Clone the repository if you have not already done so, then install the locked runtime dependencies:

git clone https://github.com/tradermonty/claude-trading-skills.git
cd claude-trading-skills
uv sync --locked
mkdir -p reports/first-week state/first-week-theses first-week-inputs

Keep every command in this guide at the repository root. We use uv run python consistently so requests, PyYAML, and jsonschema come from the locked project environment.

Day 2 — Let the Navigator Choose the Path

Ask the deterministic Navigator for a beginner-friendly 15-minute routine:

uv run python skills/trading-skills-navigator/scripts/recommend.py \
  --query "I want a 15-minute daily market check without paid API keys" \
  --no-api \
  --time-budget 15m \
  --experience beginner \
  --format json

Confirm these fields in the JSON:

primary_workflow.id = market-regime-daily
primary_workflow.api_profile = no-api-basic
no_api_path = true

The Navigator recommends; it does not execute other skills. The market-regime-daily manifest remains the source of truth for step order and required artifacts.

Day 3 — Run the No-Paid-API Market Check

Run the two required public-data analyzers:

uv run python skills/market-breadth-analyzer/scripts/market_breadth_analyzer.py \
  --output-dir reports/first-week

uv run python skills/uptrend-analyzer/scripts/uptrend_analyzer.py \
  --output-dir reports/first-week

Select exactly one timestamped output from each analyzer. The breadth pattern deliberately excludes market_breadth_history.json.

breadth_json="$(find reports/first-week -maxdepth 1 -type f \
  -name 'market_breadth_????-??-??_??????.json' -print | sort | tail -n 1)"
uptrend_json="$(find reports/first-week -maxdepth 1 -type f \
  -name 'uptrend_analysis_????-??-??_??????.json' -print | sort | tail -n 1)"

test -n "$breadth_json" && test -f "$breadth_json"
test -n "$uptrend_json" && test -f "$uptrend_json"

The workflow’s market-top step is optional, so the minimal path skips it. Pass only the two available artifacts to Exposure Coach:

uv run python skills/exposure-coach/scripts/calculate_exposure.py \
  --breadth "$breadth_json" \
  --uptrend "$uptrend_json" \
  --output-dir reports/first-week

Inspect the newest exposure_posture_*.json:

exposure_json="$(find reports/first-week -maxdepth 1 -type f \
  -name 'exposure_posture_????-??-??_??????.json' -print | sort | tail -n 1)"
test -n "$exposure_json" && test -f "$exposure_json"

uv run python - "$exposure_json" <<'PY'
import json
import sys

with open(sys.argv[1], encoding="utf-8") as report_file:
    report = json.load(report_file)

for key in (
    "inputs_provided",
    "inputs_missing",
    "confidence",
    "recommendation",
    "exposure_ceiling_pct",
):
    print(f"{key}: {report[key]}")
PY

Expect breadth and uptrend in inputs_provided, the omitted dimensions in inputs_missing, and confidence: LOW. With the critical regime and top-risk inputs missing, the fail-safe recommendation is REDUCE_ONLY or CASH_PRIORITY, never NEW_ENTRY_ALLOWED. This is a degraded, incomplete-input posture, not a complete market verdict and not evidence that the market is bearish. Optional enhancements may have additional data requirements; check their current skill guides before using them.

Day 4 — Create Your First Journal Entry

Create a manual IDEA. Use a real ticker and a falsifiable statement, but do not invent an entry or position just to complete the tutorial.

{
  "ticker": "AMD",
  "thesis_statement": "Observe whether AMD holds above the prior breakout area for five sessions.",
  "thesis_type": "growth_momentum"
}

Save that JSON and ingest it:

cat > first-week-inputs/manual-idea.json <<'JSON'
{
  "ticker": "AMD",
  "thesis_statement": "Observe whether AMD holds above the prior breakout area for five sessions.",
  "thesis_type": "growth_momentum"
}
JSON

uv run python skills/trader-memory-core/scripts/trader_memory_cli.py ingest \
  --source manual \
  --input first-week-inputs/manual-idea.json \
  --state-dir state/first-week-theses

The command prints the generated thesis ID and creates an IDEA. It does not mark a trade active and does not place an order.

Day 5 — Read the Journal Before Changing It

List the state you actually recorded:

uv run python skills/trader-memory-core/scripts/trader_memory_cli.py store \
  --state-dir state/first-week-theses \
  list

uv run python skills/trader-memory-core/scripts/trader_memory_cli.py review \
  --state-dir state/first-week-theses \
  review-due

Check that the ticker, thesis type, and status match your input. Leave the thesis as IDEA until you have separately validated a setup. Never edit the YAML state files by hand; the CLI preserves schema and lifecycle checks.

Day 6 — Turn the Steps into a Routine

Repeat Day 3 before considering new swing-trade risk. Keep this short checklist:

  1. Read both analyzer freshness warnings.
  2. Confirm which inputs Exposure Coach actually accepted.
  3. Treat missing inputs as missing; do not fill them with assumptions.
  4. Record the posture and your reasoning, not a prediction.
  5. Make every order and risk decision outside this workflow under your own rules.

The daily output is useful when it changes your process: reduce research when the posture is restrictive, or proceed to separate stock-level analysis when a complete review allows it. It is never a standalone signal.

Day 7 — Run Your First Weekly Review

Generate the trailing-seven-day digest from the local journal:

uv run python skills/weekly-performance-digest/scripts/generate_weekly_digest.py \
  --state-dir state/first-week-theses \
  --output-dir reports/first-week \
  --verbose

If you closed no trades, a zero-trade report is the correct result. Do not create a fake trade to populate the metrics. Read the generated weekly_digest_*.md and answer:

  1. Did I run the market check before considering risk?
  2. Did I distinguish missing inputs from neutral signals?
  3. Did I write a falsifiable thesis instead of a story?
  4. What one process rule will I keep or change next week?

You have now completed the smallest Plan → Record → Review → Improve loop without a paid data API. Continue with Find Your Workflow only after this routine feels repeatable.