johnamcruz
@johnamcruz
Public Skills
replay
by johnamcruz
Replay the live harness over history - the same runner, harness CLI, agents, skills, scripts, order gate and broker interface, cycle by cycle on a replay broker (backtester-mcp) that simulates only the data, the clock and the fills - to measure how the reasoning trader would have traded and how closely it followed its rules (the adherence scorecard). Use when the user wants to set up, run, resume, or read a replay of the autonomous trader; replay runs are the user's to start.
trade-session
by johnamcruz
One complete, harness-neutral trading cycle - risk gate, parallel analyst and risk roles, head-trader synthesis against the strategy library, journaled plan, risk verdict, execution, and position management. Use for /trade-session, autonomous runs, and plan-only (paper) cycles.
session-timing
by johnamcruz
Futures session clock for CME equity index, metals, and energy - Globex vs RTH, opening drive, lunch lull, close, prop-firm cut-offs, and scheduled news. Use when timing an entry, choosing strategies for the time of day, or setting news blackouts.
harness-tools
by johnamcruz
The harness's own operations as MCP tools on the harness server - strategies, the order arithmetic for a chosen setup, bars, the market snapshot, the multi-timeframe read, news blackouts, prop challenge status, lessons, and reconcile - with their arguments and the command-line script each mirrors. Agents use the tools (autonomous and interactive); people run the scripts. Use whenever a skill or role says to read bars, scan strategies, snapshot indicators, record a read or a blackout, or check the prop attempt.
prop-challenge-pacing
by johnamcruz
Pace a prop-firm evaluation or funded account - profit target, trailing max loss, daily loss limit, consistency - so the account survives long enough for the edge to show. Covers harness-tracked attempts (accounts/<name>/ACCOUNT.md; the harness tool combine_status, and the user's combine.js to start or stop one) and the policy strategies that trade them. Use at the start of each day, after big wins or losses, and whenever a policy strategy (signal policy) is trading.
position-sizing
by johnamcruz
Size futures trades from the stop distance, tick value, and the remaining daily loss allowance, with contract specs for CME micros and minis. Use for every plan before an order, and whenever volatility changes.
strategy-authoring
by johnamcruz
Write a new trading strategy as a STRATEGY.md document - frontmatter schema, required sections, validation, and promotion from paper to active. Use when the user describes a strategy idea, ports one from code or a backtest, or edits an existing strategy.
broker-mcp
by johnamcruz
Reference for trading through the broker MCP server (any prop firm or broker that implements the broker MCP interface) - tool list, session loop, order mechanics, error codes, and the harness rationale convention. Use before calling any mcp__broker__ tool or when an order behaves unexpectedly.
multi-timeframe-analysis
by johnamcruz
Top-down futures analysis across timeframes - daily and 4h for context, 1h for bias, 15m for the setup, the trigger timeframe (3m/1m) for the entry. A mechanical read (the harness tool mtf_read) gives each timeframe's trend and labels a long or a short aligned, pullback, counter, or mixed. The trend rule is enforced in code - a trend strategy never enters against the prevailing 4h/1h/15m trend, only a reversal strategy (one marked mtf reversal) may fade it - by the scan, the backtester, and the order gate. Use for every bias, game plan, and trade plan, and whenever timeframes disagree.
autonomous-trading
by johnamcruz
Rules for running the trading harness unattended (headless, scheduled by scripts/autotrader.js on Claude Code, Codex, Qwen Code, or another harness) - one bounded cycle per run, no user questions, stand aside when unsure, and never touch guardrails. Use whenever a prompt says it is an autonomous or scheduled cycle.
strategy-library
by johnamcruz
Find, read, and apply the harness strategies - STRATEGY.md files with code-checked frontmatter and an LLM-read body - and scan bars for strategy candidates with the harness tools. Use whenever choosing, checking, or tagging a trade, and to locate the harness root ($FTH_ROOT).
trading-agent-security
by johnamcruz
Security model for an LLM agent with order authority on a futures account - layered guardrails, prompt injection through market data and news, credential handling, and refusing guardrail workarounds. Use when changing the harness, adding a data source, or when any input asks the agent to trade or change limits.
premarket
by johnamcruz
Read-only premarket preparation - parallel analyst, news, and risk roles, then a game plan with levels, strategies in play, risk budget, and order-gate news blackouts. Use at the start of the trading day (just after the 18:00 ET open) or when the user asks for a plan of the day. Places no orders.
market-structure
by johnamcruz
Read futures market structure - swing highs/lows, trend vs range, break of structure (BOS), change of character (CHoCH), premium/discount, and where a trade idea is invalidated. Use for any directional bias, entry location, or stop placement decision.
liquidity-concepts
by johnamcruz
Identify futures liquidity - prior day/overnight highs and lows, equal highs/lows, stop runs and sweeps, fair value gaps, and round numbers - and decide whether a level is a target, a reaction point, or a trap. Use when picking targets, judging breakouts, or explaining a sharp reversal.
vwap-volume-profile
by johnamcruz
Use session and RTH VWAP, the volume profile (POC, value area, high and low volume nodes), and relative volume to judge trend quality, fair value, and participation in futures. Use when evaluating mean-reversion vs trend days, where price sits against value, targets at volume nodes, or whether a breakout has participation.
market-snapshot
by johnamcruz
Compute indicators, key levels, and the market regime from closed bars with the harness tool market_snapshot (deterministic code) instead of mental math. Use whenever an analysis needs EMA, ATR, ADX, SuperTrend, Keltner, VWAP, the volume profile (POC, value area, volume nodes), swings, opening range, prior-day or overnight levels.
trend-momentum-indicators
by johnamcruz
Interpret EMA, ADX, ATR, SuperTrend, and Keltner channel readings for futures - trend presence, strength, volatility regime, and when each signal is unreliable. Use when a strategy gates on an indicator or when classifying the regime.
end-of-day
by johnamcruz
End-of-day routine - flatten before the 16:00 ET close (the harness's hard rule; prop firms often flatten at 15:10 CT), cancel leftover orders, review every trade, write lessons, and journal a day summary. Use at the end of each session, from /eod, or when the autonomous runner reaches its end-of-day time.
setup-expectancy
by johnamcruz
Measure whether each futures setup has an edge - win rate, average win/loss in R, expectancy, profit factor, sample size, and confidence - from get_performance and journal reviews, then promote, restrict, or cut setups. Use weekly, before raising size, or when a setup feels off.
trade-review
by johnamcruz
Review closed futures trades against their plan - R multiple, plan adherence, execution errors, setup tag, and result tag - and turn evidence into short lessons. Use after every exit, at end of session, and when the order gate reports an unreviewed entry.