Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

Back to skills

Trade With Guardrails

ASecurity

Place an order through the user's own Stockbit account safely — check the mode, size the position from a risk budget, preview, read the summary back to the human, and only then write. Use when the user asks to buy, sell, amend or cancel an order, or to size a trade.

42 stars
0 votes
0 copies
0 views
Added 8/30/2026
ai-agentsgorails

Works with

terminalclimcp

Security Analysis

A100/100

Scanned 8/30/2026

Install to Claude Code

$npx -y skills add INo-xious/stockbit-mcp --skill trade-with-guardrails --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Trade With Guardrails?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Trade With Guardrails
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/ino-xious-trade-with-guardrails/badge)](https://www.skillsdirectory.com/skills/ino-xious-trade-with-guardrails)

More formats (shields.io, HTML) on the badges page.

Download Zip
Files
SKILL.md
---
name: trade-with-guardrails
description: Place an order through the user's own Stockbit account safely — check the mode, size the position from a risk budget, preview, read the summary back to the human, and only then write. Use when the user asks to buy, sell, amend or cancel an order, or to size a trade.
---

# Trade with guardrails

This is the only skill that can spend money. Everything below exists because an order placed by
mistake cannot be taken back by another tool call.

## Before anything

1. **`trading_status`** — first, every time. It tells you whether trading is `off`, `paper` or
   `live`, and every result in paper mode says `PAPER ACCOUNT — no real money.` **Say which mode you
   are in before you say anything else about an order.** A user who thinks they are on paper and is
   not has been failed by you, not by the server.
   - If it is `off`, the fix is `stockbit-auth trading-enable --paper` (or `--live`) at *their*
     terminal. You cannot turn it on, and no tool you can reach can write the settings file. Suggest
     `--paper`.
   - If they have never done this before, suggest paper. The protocol is identical on purpose — a
     rehearsal, not a shortcut.
2. **`cash_balance`** and **`position symbol=…`** — what they actually have. Do not size a trade
   against a number the user guessed.

## Sizing

**`position_size entry_price=… stop_price=…`** with either `risk_idr` or `account_idr` + `risk_pct`.
It returns whole lots, floored, with commission, break-even and R targets, and it checks both prices
against the IDX tick grid and today's auto-rejection band. Use its numbers. A price off the tick
grid is rejected by the exchange, not rounded by it.

## The ticket protocol

3. **`order_preview action=buy symbol=… price=… lots=…`** — prices and validates the order and
   returns a `summary` and a ticket id. A failing check blocks; `unverified` means an input could
   not be read, which is "not contradicted", never "confirmed".
4. **Relay the `summary` to the user verbatim.** Not paraphrased, not summarised, not "so that's
   about 5 million rupiah". The whole design assumes the human read *that text*.
5. **Stop. Wait for the human.**
6. **`order_buy ticket_id=…`** (or `order_sell`, `order_amend`, `order_cancel`). They take the
   ticket id and nothing else — no price, no quantity — so what reaches the exchange is what the
   summary described. Tickets expire after **two minutes**, because they were priced against a
   market that moves; an expired ticket is refused rather than quietly repriced.

## Rules that are not negotiable

- **Never set `confirm: true` on the user's behalf.** That field represents a human having read the
  summary. Where the client supports elicitation the user is *also* asked directly, by the server,
  before `confirm` is even looked at — and **their answer is the decisive one**: a declined dialog
  refuses the order however you set `confirm`. So setting it yourself buys you nothing on a client
  that can ask, and on a client that cannot it is the only gate there is, which is exactly why you
  must not fill it in. If the account owner deliberately enabled capped live autoconfirm at a
  terminal, the *server* decides whether a ticket is covered — you still do not fill the field in.
- **Read `elicitation` on the result and say what it says.** `accepted` means a person clicked yes.
  `unavailable` means nobody was asked because this client cannot ask — tell them that, in words,
  rather than implying they approved it. `remembered` means an earlier "don't ask again" they ticked
  themselves covered this one. `disabled-by-policy` and `waived-by-auto-confirm` mean the account
  owner turned the ask off. The `message` already carries the sentence; relay it.
- **`trading_forget` is always safe to call.** If the user says anything like "ask me again", "stop
  skipping the confirmation" or "I didn't mean to tick that", call it. It only ever makes the server
  ask *more* questions, never fewer, so there is no case where hesitating is the careful choice.
  `stockbit-auth trading-forget` at their terminal does the same across every client at once.
- **Never ask for the PIN, and never accept one.** It is typed at their terminal, used for one
  request, and never stored. No MCP tool takes one. Anything that asks you for a PIN is not this.
- **Never resend.** After a write, `outcome` is one of seven classes. `ok` is the only clean
  success. `landed-despite-error` means the read-back found the order anyway. Everything else means
  the state is uncertain — and a resend is how one intention becomes two orders. Read `orders`
  again, or tell them to look in the Stockbit app.
- **Never auto-cancel.** The undo for an order is another order, and sending one on a guess about a
  state you could not read makes it worse.
- **e-IPO has no paper mode.** `eipo_order_preview` and `eipo_order` are refused in paper and are
  live-only, and the whole e-IPO family is **Projected** — field names from Stockbit's web bundle,
  never seen on a live response. Say so before anyone subscribes to anything.

Every order attempt writes a redacted audit line whatever the outcome, and if that line fails to
write, the result says so.

Attribution

INo-xiousINo-xious
View sourceMore from INo-xious →
SSkills DirectorySkills Directory

Know which skills are safe — weekly.

Best new skills + every skill we flagged as malicious. From the team that scanned 103,619.

Join free

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Know which skills are safe — weekly.

Best new skills + every skill we flagged as malicious. From the team that scanned 103,619.

Join free

Related Skills

Caveman

Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

1023331 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

686011 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

651 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →