public_skills

Installable Claude Skill(s) for trading an authorised Bitfunded-style prop-firm account through the hosted webhook. Copy a skill folder into your own skills repo (or your Claude skills directory) and go.

What’s here

public_skills/
└── prop-fund-trading/          <- the skill (copy this whole folder)
    ├── SKILL.md                <- skill instructions (Claude loads this)
    ├── onboarding.md           <- how a user gets their session (Cookie-Editor)
    ├── api-reference.md        <- every webhook action documented
    ├── creds.example.json      <- credential template (copy to creds.json)
    ├── .gitignore              <- keeps creds.json out of git
    └── scripts/
        └── webhook_client.py   <- zero-dependency client (Python stdlib only)

It talks to the shared, hosted webhook: https://prop-fund-ai-trading-skill-production.up.railway.app/webhook

No server to run. The user only needs their own browser session (a cookie plus two headers). Everything is multi-user and stateless - the server stores no credentials.

Install

Into a GitHub skills repo (to share)

Copy the prop-fund-trading/ folder into your skills repository and commit it (it has its own .gitignore so a user’s creds.json never gets committed):

cp -r prop-fund-trading /path/to/your-skills-repo/skills/

Anyone who clones that repo gets the skill.

Into Claude Code

Drop prop-fund-trading/ into your Claude Code skills directory (e.g. ~/.claude/skills/prop-fund-trading/). Claude discovers it via SKILL.md.

Into the cloud (claude.ai)

Add the skill to your project / skills there. The user supplies their session in creds.json (or as PROP_FUND_* environment values); the bundled client runs in the code-execution sandbox, which has the network access it needs.

First run

1. Follow prop-fund-trading/onboarding.md to get the cookie (Cookie-Editor) and the two headers, and save them in creds.json.
2. Verify:

   cd prop-fund-trading
   python scripts/webhook_client.py status
   python scripts/webhook_client.py accounts
   

3. Ask Claude things like “show my balance”, “what’s 2% of my account on BTC”, or “open a BTC long, $50, 4x, TP 3%, SL 2%” - it drives the client for you, dry-run first.

Requirements

• Python 3.8+ (the client uses only the standard library - no pip install).
• A funded/authorised Bitfunded account you own.

Safety

Dry-run is the default. A live order needs the explicit --live flag. Max leverage is capped at 5x server-side, and every order must carry a take profit and a stop loss. Your creds.json holds your session - keep it private and never commit it. See prop-fund-trading/SKILL.md and api-reference.md for the full rules.

Updating

The skill calls the hosted webhook, so new server-side features appear automatically with no change on the user’s side. If the webhook URL ever