Public Agent Skills repo for TickerAll (SKILL.md packs for Claude Code, Cursor and other coding agents)

Summary Ship first-party Agent Skills for TickerAll in a public repo so AI coding agents (Claude Code, Cursor, Codex, and any agent that reads SKILL.md folders) know how to build against TickerAll correctly, without pasting the docs into every session. Today an agent asked to "connect my MT5 account and stream ticks through TickerAll" has to discover the REST/WS surface, the auth model, the tier rules and the two SDKs from scratch, and usually guesses something wrong: base URL, auth header, checking status instead of hot, polling instead of the WS stream, server time vs UTC on candles. A skill fixes that once, for every agent. Where it lives A new PUBLIC repository (working name TickerAll/skills), MIT-licensed, alongside the public SDKs. Users clone it or copy a folder into their agent's skills directory. Nothing in it references anything below the public API surface, same rule as the SDK READMEs and the docs. Proposed skills (one folder each, SKILL.md with name + description frontmatter, optional references/ and scripts/) - tickerall-api: auth (Authorization: Bearer cf_api_*), base URL, account lifecycle (connect -> hot -> trade/read), error codes, tier limits, rate and idempotency conventions - tickerall-sdk-ts: @tickerall/sdk install and idiomatic usage: accounts, positions, place/close/modify, candles, history, WS tick stream with reconnect - tickerall-sdk-python: same for the tickerall PyPI package (sync + async) - tickerall-mcp: adding mcp.tickerall.com as a connector, the 15 tools, when to prefer MCP over the SDK - tickerall-market-data: candles (UTC timestamps vs terminal server time), tick stream, symbol specs, deep history - tickerall-trading: place/close/modify semantics, demo-vs-live checks, SL/TP on market orders, copy-trading basics, safe defaults tickerall-api is the minimum first cut; the rest can land incrementally. Deliverables 1. Create the public repo with license, README, CHANGELOG and CI that lints every SKILL.md and runs smoke scripts against staging with a CI-only key 2. One folder per skill above, grounded in the live docs and SDK source, with runnable snippets tested against staging 3. scripts/ helpers where useful (e.g. a smoke script that lists accounts and prints the first tick) 4. Install instructions per agent: Claude Code (~/.claude/skills/ or project .claude/skills/), Cursor rules, plain copy for other agents 5. Link from /docs and the homepage next to the MCP section ("Works with your AI coding agent") 6. Add the skills repo to the release checklist: an API change bumps @tickerall/sdk, the tickerall PyPI package, MCP, docs AND the skills Acceptance A fresh Claude Code session with the skills installed and only an API key can, without further prompting: connect a demo account, wait for hot, place and close a demo trade, fetch H1 candles with correct UTC handling, and stream ticks over WS. Why now The MCP server already dropped the integration bar for chat-style use. Skills do the same for code-writing agents, which is how most bot developers and platform builders actually work. A public skills repo is also a discovery surface on its own (GitHub search, agent skill directories).