Astra Dual-Signal Intraday Auto-Trader

    1

    The complete operating rulebook of **Astra**, HPSILab's autonomous intraday trading agent, packaged so your AI agent can **build, audit, operate or backtest** it.

    $1000

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Astra Dual-Signal Intraday Auto-Trader

    Example session with this skill installed

    Use the Astra skill. My bot holds these intraday positions, each bought at $100.00 at 10:05 ET. For each one, tell me what the exit engine does right now and whether I may buy the symbol again today:
    A) last $96.00 at 10:25
    B) last $94.90 at 10:25
    C) last $103.20 at 10:25
    D) last $97.90 at 10:36
    E) last $103.10 at 10:50
    F) last $100.60 at 10:45, both Signal_14 and Signal_5 say Sell
    G) went to $102.10 at 10:40, now $99.95 at 10:55
    H) last $97.50 at 15:50, dual Sell active

    • Read your context and instructions
    • Compiled the astra dual-signal intraday

    Summary

    Ran each position through scripts/exit_rules.py (rulebook v1.4.2, §5). Cost $100.00, first fill 10:05 ET, EOD liquidation at 15:50. Six positions exit now, two hold, and three symbols are blocked for the rest of the session.

    Decisions

    PosHeldLastStageDecisionConditions firedRe-buy today?
    A20 min$96.00 (-4.0%)ProtectionHOLDnonen/a
    B20 min$94.90 (-5.1%)ProtectionHARD_STOPHARD_STOPNo (stop)
    C20 min$103.20 (+3.2%)ProtectionHOLDnone (TP inactive before 30 min)n/a
    D31 min$97.90 (-2.1%)NormalSTOP_LOSSSTOP_LOSSNo (stop)
    E45 min$103.10 (+3.1%)NormalTAKE_PROFITTAKE_PROFITYes, on a new dual Buy
    F40 min$100.60 (+0.6%)NormalSIGNAL_SELLSIGNAL_SELLYes, on a new dual Buy
    G50 min$99.95 (-0.05%)Normal, breakeven armed at 10:40 (lock $100.00)BREAKEVENBREAKEVENYes, on a new dual Buy
    H345 min$97.50 (-2.5%)NormalEND_OF_DAYEND_OF_DAY, STOP_LOSS, SIGNAL_SELLNo (stop fact recorded)

    Why

    A and C hold because the first 30 minutes only allow the -5% hard stop (plus EOD and emergency). A -4% dip or a +3% spike in this window is not an exit, and the -4% is not recorded as a stop.

    B crossed the hard stop at $95.00. The stop is written on first detection with blocked_session = today, whether or not the sell fills.

    • D is past 30 minutes and below the $98.00 stop with no breakeven armed.

    G armed breakeven when it reached $102.00 after 30 minutes; the effective stop moved up to $100.00 and never moves down. Exiting there is BREAKEVEN, which does

    not block re-entry.

    H fires three conditions at once. Priority is emergency > hard stop > abnormal carry > EOD > stop loss > breakeven > take profit > dual sell, so one sell goes out tagged END_OF_DAY, while the STOP_LOSS fact is still stored and blocks the symbol.

    Next steps

    1. Persist each decision (exit intent, reason, all fired conditions)

    before sending one sell per position for the confirmed remaining quantity.
    2. Use a deterministic idempotency key per order; if the result is unknown, query by key instead of resending.
    3. Keep B, D and H on today's no-rebuy list; clear it at the next session.
    4. Run python scripts/test_skill.py (23 tests) after adapting thresholds to your config.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The complete operating rulebook of Astra, HPSILab's autonomous intraday trading agent, packaged so your AI agent can build, audit, operate or backtest it. Version 1.4.2, written and reviewed as an engineering spec, not a tip sheet.

    What your agent gets

    • Entry rules: a fresh dual buy (RSI(14) and RSI(5) signals with MACD and Bollinger) on completed 1-minute bars only. No half-formed bars, no replays after a restart, one execution per event.
    • Exit engine with a strict priority order: a 30-minute protection period with a -5% hard stop, then +3% take profit, -2% stop loss, a breakeven lock that never moves down, dual sell signal and end-of-day liquidation.
    • Risk gates: 10% sizing, position, sector and exposure caps, open stop risk, daily entry cap, -3% daily loss halt, drawdown lock, no same-day re-entry after a stop.
    • Owner kill switch for event days (FOMC, CPI, earnings): blocks every buy path, never blocks an exit, fails closed.
    • Execution safety: idempotent order keys, commit boundaries, crash recovery, reconciliation, run modes and a manual-review path.
    • Data model for events, trades, order attempts, fills, baselines, equity marks and halts.
    • Acceptance checklist of about 80 test cases.
    • Reference Python: signals, a pure exit-decision function and a minute-bar backtester with a random-entry control. 23 self-tests included.

    Who it is for

    Developers and quant builders who want an AI agent to implement or review a disciplined intraday system, add proper exits and recovery to an existing bot, or test the rules on their own data.

    What it is not

    Not a signal service, not a promise of returns and not financial advice. The rules are a framework that must be measured after realistic costs. The skill instructs your agent to work on paper first and never touch a live account without your explicit approval.

    US stocks, regular session only. Needs 1-minute OHLCV data from your own provider. No broker integration included. The backtester simulates one position per symbol and does not model portfolio limits. Rules are not a validated edge; results depend on your data and costs.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 11 days ago

    Permissions

    Terminal / Shell
    Read Files
    Network Access
    Listed11 days ago

    What's inside

    Frequently Asked Questions

    No. It is a disciplined rule framework for building, auditing and testing an intraday trading agent, not a signal service and not financial advice. The included backtest guide tells your agent to measure the rules after realistic costs against controls before trusting them.

    The skill is broker-agnostic and ships no broker code. It defines the order lifecycle, idempotency keys and reconciliation your adapter must follow. It instructs your agent to run on a paper account first and never place live orders without your explicit approval.

    SKILL.md; five references (full rulebook, state machine and recovery, data model, acceptance checklist, backtest method); four Python scripts (signals, exit decision engine, minute-bar backtester, 23 self-tests); and a license.

    The rulebook persists every intent before the order, uses deterministic order keys and defines recovery for each commit step: query the execution side by key, never resubmit blindly, re-run all entry checks before retrying a buy, and hand any position whose state cannot be rebuilt to manual review.

    It is designed for US stocks during regular NYSE hours (entry window, end-of-day exit, session baselines). Adapting it to 24/7 markets would require redefining sessions and is not covered.

    Yes. Every number (stops, targets, sizing, limits, windows, signal thresholds) is a configuration value listed in the rulebook, and the scripts take them as parameters. The defaults are the reviewed design values, not optimized ones.

    New in Trading & Investing