Agent LaneGuides
All Agent Lane guides
97 plain-language guides for people who let AI agents touch wallets, brokerages, and paid APIs.
Tools
- AI agent spend cap calculator
Per-payment, daily and monthly caps, a rate limit and an approval threshold from your agents’ usage.
AI scams and your money
- AI crypto scams: how they work and how to protect your crypto
Fake support agents, fake AI trading bots, AI-written phishing, wallet-drainer links and pig-butchering chats: how each works, the red flag, and the rules that protect your crypto. - Is that AI trading bot legit? How to check before you hand over money or API keys
Red flags, guaranteed-return claims, why a small test withdrawal proves little, and read-only vs trade vs withdraw API key permissions. - Check a call or video before you send money: voice clones and deepfakes
How voice-clone and deepfake video money requests work, hang up and call back, a family safe word, and a short checklist before you send money or crypto.
Start here
- Dedicated agent wallet
Keep the bot’s spend wallet separate from long-term holdings — fund only what one bad run can lose, revoke and rotate before you raise the balance, hardware cold store vs hot agent wallet. Publisher checklist; Agent Lane does not hold keys. - Agent spend caps
Stack monthly budget, session ceiling, and daily max so a live bot cannot burn the plan. Operator checklist · Agent Lane. - Review before place
Agent proposes; human confirms; then place — the human-in-the-loop gate before an agent can spend, trade, or move money. Pairs with spend caps and a dedicated wallet. Publisher checklist; Agent Lane does not place trades. - Shield brokerage spend
Protect the brokerage MCP trading path before an agent can place — dedicated funded Agentic sleeve, spend caps, review before place; no “just place it.” Publisher checklist; Agent Lane does not place trades. - Shield crypto spend
Protect the crypto spend path before an agent can sign or send — dedicated hot wallet, spend caps, review before place, revoke and rotate; cold store stays offline. Publisher checklist; Agent Lane does not hold keys. - x402 buyer basics
How a person or agent pays for a web request with x402 under spend limits — dedicated wallet, caps before any signature, revoke before debug. Educational guide; Agent Lane does not hold keys. - Trading-agent map — brokerage, crypto, prediction lanes
Map which bot or path touches brokerage MCP, crypto wallet, prediction markets, or research-only. Match the shield page. Never skip dedicated wallet, spend caps, and review before place. Specialist bots, not one mega-bot. - Shared-box and connector checklist
Lock down a shared computer, shared logins, and Cursor/Grok connectors before any agent spend. Map who can sit at the box, which passwords the agent can reach, and how far one leak can spend. Publisher checklist; Agent Lane does not audit your machine. - Install a specialist bot, don’t paste a mega-prompt
Specialist crew beats one mega-bot: one clear job per marketplace bot, connect only what that job needs, wire private vs shared memory, keep spend behind a dedicated wallet, caps, and review before place. - Agent memory hygiene
Private desk scratch stays with one bot; house facts go in shared user-memory. Specialist crew beats one mega-bot. Prune weekly. Memory never replaces review before place. Publisher checklist; Agent Lane does not edit your memories.
Session and order rules (A–Z)
- A proactive ping is not permission to spend
An agent that offers to book a ride or a flight has not been told to pay. A proactive message is not a booking and not a charge. Pair it with spend caps and a session ceiling. Publisher checklist; Agent Lane does not hold funds or place trades. - Cap daily loss before session
Before an agent money session starts, set a hard daily (or session) loss cap and stop rules — so a losing streak cannot keep placing after the budget for the day is gone. Pair with fund-session-ceiling, session-time-box, name-the-strategy-before-connect, never-skip-review-before-place, and two-person-rule-for-large-places. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Confirm buying power before place
Before the agent places, confirm buying power / available cash in the broker (not chat math) so you don’t submit into a reject, partial, or margin surprise. Pairs with never-skip-review-before-place, confirm-fills-in-broker-not-chat, read-only-quote-before-trade-connector, dry-run-before-first-live-place, size-from-sleeve-not-from-hope. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Confirm cancel in the broker, not chat
When an agent says it cancelled an order or closed a position, verify the cancel or flat state in the broker or wallet UI yourself — chat is not the source of truth. Pairs with confirm-fills-in-broker-not-chat, never-skip-review-before-place, log-session-end-checklist, journal-every-override. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Confirm fills in the broker, not chat
Treat the broker or wallet UI as the source of truth for fills and balances — chat summaries and agent claims can lag, hallucinate, or miss partials; confirm there before the next place or before you raise the sleeve. Publisher checklist; Agent Lane does not place trades or hold funds. - Confirm symbol before place
Before place, human confirms the exact symbol/pair/contract in the broker UI matches the written thesis — reject lookalike tickers, wrong venue listings, and agent chat typos. Pairs with reject-agent-place-without-quote, confirm-buying-power-before-place, never-skip-review-before-place, confirm-fills-in-broker-not-chat, one-venue-one-agent-session, write-invalidation-before-entry. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Connect only what you need
Before a live agent run, enable only the MCP connectors and tools required for that job — an unused brokerage, payer, email, or browser connector still expands blast radius if the prompt drifts; code connectors vs money connectors, with spend caps and a dedicated wallet for any job that spends. Publisher checklist; Agent Lane does not install connectors for you. - Dry run before first live place
Before the first live place with a new agent, connector, or strategy, run a dry session (paper, preview, or zero-size rehearsal) and confirm fills in the broker UI — pair with paper-before-live, never-skip-review-before-place, and confirm-fills-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades or hold funds. - Empty the sleeve overnight
Do not leave spendable cash on a dedicated agent wallet or Agentic sleeve beside live connectors overnight (or across a long idle) — empty or park excess when the session ends, keep review-before-place on, and only re-fund the session ceiling when you are back at the desk. Publisher checklist; Agent Lane does not hold funds or place trades. - Flatten all before connector revoke
Before you revoke or disconnect a live trading or spend connector, flatten or cancel working risk in the broker UI first — revoke without flattening can leave orphan working orders or open exposure you can no longer manage through the agent path. Pairs with revoke-before-debug, revoke-connectors-when-idle, confirm-cancel-in-broker-not-chat, one-cancel-one-confirm, empty-sleeve-overnight, kill-switch-before-travel. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Flatten before news you cannot size
Before binary / high-impact news you have no written edge for (FOMC, Clarity-style crypto bills, earnings you don’t trade), default flatten or cut size — do not hold overnight leverage into unknown catalysts. Pairs with no-overnight-margin-without-go, empty-sleeve-overnight, kill-switch-before-travel, cap-daily-loss-before-session, session-time-box. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Fund the session ceiling
A monthly budget plan is not the cash you load for one session. Fund only the session ceiling you already set, and park any surplus the agent can reach. Publisher checklist; Agent Lane does not hold funds or place trades. - Give your AI agent its own email inbox
Why an AI agent should have its own email address instead of yours, what it can sign up for with it, what still needs a human, and how to keep spend and safety in check. - Robinhood MCP: what an AI agent can and can’t do
What the official Robinhood Trading MCP is, what a connected AI agent can read and trade, what it can’t do, and the safety controls and limits to set first. - Virtual cards for AI agents: what limits actually exist
Nine card options for AI agents compared on spend limits, merchant locks, single-use cards, real-time approval, who can sign up, and cost. - Can AI agents make purchases? How it works and how to set limits
Yes, AI agents can buy things once you give them a way to pay. How agents pay, what goes wrong, and the limits to set first. - AI agent sent crypto by mistake: what you can and can’t undo
What a confirmed transfer means, the first-hour steps (stop, revoke keys and approvals, call the exchange), recovery-scam warnings, and the limits that keep it small. - Should I give my AI agent my credit card? Safer ways to let it pay
Do not hand an AI agent your main credit card. Use a separate virtual card with a hard limit, merchant locks, and an ask-first rule. What can go wrong and a 6-step safer setup. - How to set up Grok Bot safely
A plain-language setup for Grok Bot: its own inbox, you approve every send, a capped card instead of your main one, read-only first, stop conditions for recurring jobs, and an honest look at whether the X Subscription is worth it. - What is an AI agent? A plain-language guide
An AI agent is software that takes a goal, picks its own next steps, and uses tools to act for you. - Agentic commerce explained: what it means when AI agents shop for you
Agentic commerce means AI agents finding, choosing, and paying for things for you. - An AI agent with a wallet or email inbox: what it can do and how to limit it
An AI agent with its own wallet can pay and receive money; with its own inbox it can sign up, receive codes, and send mail. - AI agent tools vs skills vs MCP: what each one is and which to turn on
Tools let an AI agent act, skills tell it how to do a kind of job, and MCP is the plug that connects outside apps. - How much does an AI agent cost per month? The real costs and how to cap them
An AI agent costs a subscription or usage fees, plus the tools it calls and whatever it buys for you. - Best AI agent for small business: how to choose by job, not by hype
There is no single best AI agent for a small business. - How to keep an AI agent secure: a 9-step checklist
An AI agent is only as safe as the access you give it. - What is prompt injection? How hidden instructions trick AI agents
Prompt injection is when text an AI agent reads contains instructions meant to take over what it does. - AI agent vs automation: which one does your business need?
Automation follows a fixed recipe you wrote: when this happens, do that. - AI agent for customer support: what it can handle and how to set it up safely
An AI agent for customer support reads incoming questions, answers the common ones from your own help content, sorts the rest, and drafts replies for a person to check. - Can AI reply to my Google reviews? What can go wrong
What can go wrong when AI answers your Google reviews (wrong facts, copy-paste replies, private details, policy problems) and the safer setup: AI drafts, a person approves. - How to test an AI agent before you give it real access
Before an AI agent touches your real inbox, files, customers, or money, run it somewhere a mistake costs nothing. - AI agent audit logs: what to record and how to review them
An audit log is a dated record of what an AI agent did: the request it got, the tools it used, what it sent, changed, or spent, and who approved it. - AI agent for bookkeeping: what it can do and what to keep for your accountant
An AI agent for bookkeeping does the tedious first pass: it reads receipts and bank lines, suggests a category for each, matches payments to invoices, and lists what is missing. - AI agent for scheduling: how to let it book meetings without double-booking you
An AI agent for scheduling finds open times on your calendar, offers them to the other person, books the meeting, and sends reminders. - AI agent for sales outreach: how to use it without spamming or burning your domain
An AI agent for sales outreach researches who to contact, drafts a short personal message for each, and keeps track of follow-ups. - Multi-agent systems explained: when several AI agents beat one
A multi-agent system is a setup where several AI agents work on one job, each with its own role: one researches, one writes, one checks, and often one coordinates the others. - AI agent memory explained: what agents remember and how to control it
AI agent memory is how an agent keeps information it can use later: what you said earlier in a task, your preferences from last week, or facts about your business. - How to choose an AI agent platform: 8 things to check before you sign up
An AI agent platform is the product your agents run on: it supplies the model, the tools, the connections to your apps, and the rules. - Which AI agent actions need a human OK? A simple approval rule
If you cannot undo it in about a minute, the agent asks first: what can run alone and what needs your OK. - How to run a one-person company with AI bots, safely
Give each bot one job, write the rules down in files you own, and keep every send, purchase, and delete behind your yes. - Free AI agent courses for small-business owners
13 free courses and official guides for learning to run AI agents, grouped into start here, build agents, and run them safely. - Does Robinhood have an AI trading bot? What actually exists in 2026
Robinhood does not sell a “set it and forget it” trading bot. It offers built-in Robinhood Agents and an official MCP connector for your own AI agent, both trading in a separate account. What each does and the limits to set. - How AI agents sign in to apps with AgentID
AgentID is like “Sign in with Google,” but for AI agents: your agent signs in to apps with its own AgentMail inbox instead of your login. What it is, why the agent needs its own inbox, how sign-in works step by step, and what to watch: what the app learns, which inbox owns the account, how to revoke, and spend. - Journal every override
Whenever a human overrides the agent’s plan, skips a review gate, or changes size/kill rules mid-session, write it down — why, what changed, what filled — so the next session inherits reality, not hope. Pairs with log-session-end-checklist, receipt-log-after-session, never-skip-review-before-place, name-the-strategy-before-connect, no-averaging-down-without-new-thesis. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Kill switch before travel
Before travel or extended offline time, pause or revoke agent spend access (empty sleeve / revoke connectors / lower ceilings) so a stuck session cannot place while you cannot review — pair with empty-sleeve-overnight, revoke-connectors-when-idle, fund-session-ceiling, and session-time-box. Publisher checklist; Agent Lane does not place trades or invent broker UI. - Log session end checklist
End every agent money session with a short written checklist — fills confirmed in the broker UI, receipts filed, connectors still needed or revoked, sleeve/ceiling state noted. Pair with receipt-log-after-session, confirm-fills-in-broker-not-chat, revoke-connectors-when-idle, and empty-sleeve-overnight. Publisher checklist; Agent Lane does not place trades or invent broker UI. - Max open positions per sleeve
Cap how many simultaneous open positions a sleeve (or named strategy) may hold so an agent cannot stack risk across too many names at once. Pairs with one-strategy-one-sleeve, fund-session-ceiling, cap-daily-loss-before-session, name-the-strategy-before-connect, never-skip-review-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Max slippage or cancel
Write a max slippage / worst acceptable fill before place — if the print would breach it, cancel rather than chase. Pairs with no-market-order-without-go, reject-agent-place-without-quote, never-skip-review-before-place, size-from-sleeve-not-from-hope, confirm-cancel-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Name the strategy before connect
Before you connect a broker, wallet, or spend connector, write down the strategy name, allowed instruments, size ceiling, and kill conditions — so the agent inherits a named plan instead of improvising. Pair with connect-only-what-you-need, one-connector-at-a-time, paper-before-live, dry-run-before-first-live-place, and never-skip-review-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Never paste API keys in chat
Never paste API keys, session cookies, seed phrases, or broker passwords into a chat with an agent — once in the transcript they can be logged, replayed, or leaked; use the product’s secret/vault path or a dedicated connector grant instead. Publisher checklist; Agent Lane does not hold credentials or run your agents. - Never share a seed phrase with an agent
Do not paste, upload, or store a wallet seed phrase, recovery phrase, or private key in agent chat, memory, connectors, or shared-box notes — fund only a dedicated spend path the agent can reach; long-term holdings stay on hardware or accounts the bot cannot export. Publisher checklist; Agent Lane does not hold keys or run wallets. - Never skip review before place
Do not grant standing “just place it / don’t ask / skip the card” permission for live brokerage or payment connectors — review-before-place is the human gate on amount, destination, asset, and venue; a one-time ask does not authorize an open-ended auto-place habit. Publisher checklist; Agent Lane does not place trades. - No averaging down without a new thesis
Do not let an agent add size to a losing position just because the price is lower — require a fresh written thesis, size check against the sleeve ceiling, and review gate before any add — pairs with name-the-strategy-before-connect, never-skip-review-before-place, cap-daily-loss-before-session, one-strategy-one-sleeve, two-person-rule-for-large-places. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No chase after partial fill
After a partial fill, do not let the agent chase the remainder with worse prices, marketables, or “just finish the size” — cancel/leave working only what the written plan allows, confirm residual in the broker UI, and require a fresh GO + quote + slippage bound before any add. Pairs with max-slippage-or-cancel, confirm-fills-in-broker-not-chat, confirm-cancel-in-broker-not-chat, no-market-order-without-go, no-averaging-down-without-new-thesis, reject-agent-place-without-quote. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No market order without GO
Market (or marketable) orders need an explicit written GO — default is limit/last-look so the agent cannot chase a print into a gap. Pairs with reject-agent-place-without-quote, never-skip-review-before-place, size-from-sleeve-not-from-hope, confirm-buying-power-before-place, write-go-before-scale-up. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No new risk after session end
When the written session clock ends (or you hit the time-box), stop opening new risk — flat or hold only what the overnight/GO rules already allow; do not “one more trade” past the bell. Pairs with session-time-box, log-session-end-checklist, empty-sleeve-overnight, no-overnight-margin-without-go, stop-adding-size-after-loss-day. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No overnight margin without GO
Do not let an agent leave leveraged or margin exposure open overnight (or across a session you will not watch) unless you explicitly GO that risk — default is flatten or reduce before sleep. Pairs with empty-sleeve-overnight, kill-switch-before-travel, session-time-box, cap-daily-loss-before-session, confirm-cancel-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No place while unconfirmed fill
Do not let the agent place a new order while a prior fill, cancel, or flat is still unconfirmed in the broker UI — wait for venue truth before the next place. Pairs with confirm-fills-in-broker-not-chat, one-cancel-one-confirm, no-chase-after-partial-fill, one-order-per-idea-until-fill, never-skip-review-before-place, confirm-buying-power-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - No scale-in without GO
Do not scale into a winning (or open) position without a written GO that sets add size, max total risk, and invalidation — pyramid/scale-in is not free just because price moved your way. Pairs with no-averaging-down-without-new-thesis, write-go-before-scale-up, one-risk-unit-per-idea, size-from-sleeve-not-from-hope, write-invalidation-before-entry, never-skip-review-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - One cancel, one confirm
For every cancel (or flatten/close) the agent claims, require one human confirm in the broker UI before the next place or size change — one cancel → one confirm; chat is not enough. Pairs with confirm-cancel-in-broker-not-chat, confirm-fills-in-broker-not-chat, no-chase-after-partial-fill, never-skip-review-before-place, max-slippage-or-cancel, journal-every-override. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - One connector at a time
After revoke/debug or a clean start, reconnect only one payer/brokerage/MCP connector at a time, with review-before-place on, funded only to the session ceiling — never re-enable the whole stack at once. Publisher checklist; Agent Lane does not hold funds or place trades. - One order per idea until fill
Keep at most one working order per thesis until it fills, cancels, or you write a replace GO — no ladder spam / multi-ticket stacking that quietly multiplies the same idea. Pairs with reject-duplicate-place-same-thesis, one-risk-unit-per-idea, no-chase-after-partial-fill, max-open-positions-per-sleeve, no-scale-in-without-go, never-skip-review-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - One risk unit per idea
Each distinct thesis gets one predefined risk unit (max loss in sleeve terms) — do not stack correlated tickets that quietly multiply the same idea’s risk. Pairs with size-from-sleeve-not-from-hope, max-open-positions-per-sleeve, one-strategy-one-sleeve, no-averaging-down-without-new-thesis, write-go-before-scale-up. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - One strategy, one sleeve
Give each named strategy its own funded sleeve (or clearly labeled sub-balance) so P&L, ceilings, and kill switches do not blur across unrelated plays. Pair with name-the-strategy-before-connect, separate-funding-from-trading-wallet, fund-session-ceiling, and cap-daily-loss-before-session. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - One venue, one agent session
Keep one agent session pointed at one trading venue (or one broker account) so fills, cancels, and kill-switches stay unambiguous — do not let the same chat juggle two brokers or mix paper + live without a hard wall. Pairs with separate-browser-profile-for-agent, one-strategy-one-sleeve, separate-funding-from-trading-wallet, never-skip-review-before-place, confirm-fills-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Paper before live
Before funding a live brokerage/payment connector or Agentic sleeve, run the same stack on paper (or a zero-funded / preview-only path): review-before-place on, one connector at a time, caps written — prove the pipe works without spendable cash, then fund only the session ceiling. Publisher checklist; Agent Lane does not hold funds or place trades. - Paper before live, same rules
Prove the sleeve’s written rules on paper (or dry-run) before live money — and keep the same size caps, loss limits, and GO gates when you go live; do not loosen rules just because cash is real. Pairs with dry-run-before-first-live-place, write-go-before-scale-up, one-venue-one-agent-session, size-from-sleeve-not-from-hope, name-the-strategy-before-connect. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Read-only quote before trade connector
Connect market-data / quote access first (read-only) and prove the agent can see prices cleanly before you grant place, trade, or spend permissions — pairs with one-connector-at-a-time, connect-only-what-you-need, paper-before-live, dry-run-before-first-live-place, and name-the-strategy-before-connect. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Receipt log after the session
After every live agent spend or place session, write down what ran (amount, destination, asset, venue, connector, time) before you empty the sleeve or raise caps — a short receipt log beats memory when something looks wrong later. Publisher checklist; Agent Lane does not place trades or hold funds. - Reject agent place without a fresh quote
Do not let the agent place on a stale chat number — require a fresh broker quote (or last-look) in the same review step as size and buying power. Pairs with read-only-quote-before-trade-connector, confirm-buying-power-before-place, never-skip-review-before-place, dry-run-before-first-live-place, confirm-fills-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Reject duplicate place same thesis
Reject a second (or nth) place that is the same thesis while a working order or open risk for that idea already exists — one risk unit, one live ticket path; no accidental double-submit from chat retries or agent loops. Pairs with one-risk-unit-per-idea, no-averaging-down-without-new-thesis, max-open-positions-per-sleeve, no-chase-after-partial-fill, never-skip-review-before-place, write-invalidation-before-entry, confirm-fills-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Revoke before you debug
Disconnect, lock, empty the sleeve, revoke or rotate — then dig into the bad prompt. The incident checklist before you debug a live agent spend path. Pairs with review-before-place, spend caps, and a dedicated wallet. Publisher checklist; Agent Lane does not hold funds or place trades. - Revoke connectors when idle
When you are not actively running an agent session, revoke or disconnect spend and trading connectors so a stale grant cannot place while you are away — pair with empty-sleeve-overnight and rotate-before-raise-balance. Publisher checklist; Agent Lane does not place trades or hold funds. - Rotate before you raise the balance
Before you raise the funded ceiling on a dedicated agent wallet or Agentic sleeve, revoke/rotate the live connector credentials (and confirm review-before-place still on) — do not top up a pipe you have not re-verified after a bad prompt, share, or long idle. Publisher checklist; Agent Lane does not hold funds or place trades. - Separate browser profile for the agent
Run agent browser sessions in a dedicated profile (or computer) so personal banking cookies, password vault unlocks, and long-lived sessions do not sit beside live spend connectors — pairs with shared-box hygiene and revoke-before-debug. Publisher checklist; Agent Lane does not run browsers for you. - Separate funding from trading wallet
Keep long-term, cold, and funding capital off the hot wallet or broker sleeve the agent can spend — only move what the session ceiling allows. Pair with fund-session-ceiling, empty-sleeve-overnight, never-share-seed-with-agent, and kill-switch-before-travel. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker/wallet UI. - Session time-box
Cap how long an agent trading or spend session may run before a hard stop — time-boxing limits blast radius when a loop or bad prompt keeps placing. Pair with fund-session-ceiling and empty-sleeve-overnight. Publisher checklist; Agent Lane does not place trades or hold funds. - Size from sleeve not from hope
Position size comes from the sleeve’s written risk rules — max % of sleeve, max loss, open-position cap — not from conviction, FOMO, or “it’ll come back.” Pairs with one-strategy-one-sleeve, max-open-positions-per-sleeve, cap-daily-loss-before-session, no-averaging-down-without-new-thesis, no-overnight-margin-without-go. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Stop adding size after a loss day
After a session that hits the sleeve’s daily loss (or a defined loss streak), do not add size or open new risk until a written reset GO — revenge sizing and “make it back” are the failure mode. Pairs with cap-daily-loss-before-session, size-from-sleeve-not-from-hope, no-averaging-down-without-new-thesis, write-go-before-scale-up, journal-every-override. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Two-person rule for large places
For places above a size you set, require a second human review (or a hard Auto-review gate you never bypass) before the agent can submit — pair with never-skip-review-before-place, fund-session-ceiling, dry-run-before-first-live-place, and confirm-fills-in-broker-not-chat. Publisher checklist; Agent Lane does not place trades or invent broker UI — you define the threshold. - Whitelist merchants only
If your agent can spend, restrict it to an allowlisted set of merchants or counterparties so a prompt or loop cannot pay an unexpected payee — pair with fund-session-ceiling, connect-only-what-you-need, and never-skip-review-before-place. Publisher checklist; Agent Lane does not place trades or hold funds. - Write GO before scale-up
Before you raise size, add leverage, open a second strategy, or widen the sleeve, write an explicit GO (what changes, max loss, kill criteria) — do not scale on a green day alone. Pairs with size-from-sleeve-not-from-hope, name-the-strategy-before-connect, dry-run-before-first-live-place, two-person-rule-for-large-places, journal-every-override. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Write invalidation before entry
Before the agent enters a live ticket, write what would invalidate the thesis — price, time, structure, or catalyst fail — so exit is pre-committed, not improvised after you are in. Pairs with name-the-strategy-before-connect, one-risk-unit-per-idea, no-averaging-down-without-new-thesis, flatten-before-news-unknown, never-skip-review-before-place, size-from-sleeve-not-from-hope. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI. - Write max loss before entry
Before any live place, write the max loss for that ticket (dollars or % of sleeve) in the review card / plan — so exit size and stop logic are pre-committed; if the place would risk more than that written max, reject. Distinct from daily loss cap (session budget) and from invalidation (why you exit) — this is the how much you can lose gate before entry. Pairs with write-invalidation-before-entry, one-risk-unit-per-idea, size-from-sleeve-not-from-hope, cap-daily-loss-before-session, never-skip-review-before-place, confirm-buying-power-before-place. Publisher checklist; Agent Lane does not place trades, hold wallets, or invent broker UI.