--- name: cornerstones-market-data-context description: Use when a shell-capable agent needs Cornerstones market data. Installs cornerstones-client, discovers live market data commands, verifies API-key access, respects account access, and checks source labels, freshness, and data quality before using quotes, charts, order flow, events, evidence, or context packets. version: 1.0.0 author: Cornerstones license: MIT metadata: hermes: tags: [market-data, trading-agents, cli, charts, order-flow, evidence] related_skills: [] --- # Cornerstones Market Data Skill ## Overview Use this skill when an agent needs to use Cornerstones market data safely from a shell. ## When to Use - The user asks an agent to install or use Cornerstones. - The agent needs FX, equity, A-share, crypto, options, macro, chart, event, evidence, order-flow, or context-packet reads. - The agent needs to verify a Cornerstones API key and distinguish account access before consuming market data. - The agent needs a safe checklist for source labels, freshness, quality, and availability state. Do not use this skill for broker trade execution, account operations, or claims about unsupported historical/latency coverage. ## What Cornerstones is Cornerstones is a read-only market context layer for AI trading and research agents. It packages market data, chart artifacts, order flow, events, evidence, and context packets into inspectable JSON. It is not a broker, not an execution engine, and not a promise that every market surface is available on every plan. ## First rule: discover reality before assuming commands Always run live help and onboarding commands before choosing a data path: ```bash python -m pip install -U cornerstones-client cornerstones-client --help cornerstones-client guide cornerstones-client changelog cornerstones-client auth status ``` If a command shown in older docs is missing, trust `cornerstones-client --help` and the command-specific `--help` output. ## API key setup A Cornerstones API key is separate from broker, exchange, model-provider, and upstream data-provider keys. After the user provides or issues a key from the Cornerstones dashboard: ```bash cornerstones-client auth login --api-key cornerstones-client verify cornerstones-client guide ``` Do not ask for unrelated credentials. Do not paste the key into public logs. ## Account access Treat access state as part of reasoning: - Browser trial: discovery only; it is not a market-data key. - Free: basic authenticated market truth. - Pro: chart artifacts, context packets, options chain/wall/analysis, macro exact series, and premium event export when enabled. - Max: all Pro surfaces plus orderflow raw, summary, context, historical reads, and liquidity metrics. If a response says a surface is unavailable or requires a different account level, preserve that fact in the next reasoning step instead of hiding it. ## Safe market data commands to inspect Run command-specific help first: ```bash cornerstones-client fx --help cornerstones-client stocks --help cornerstones-client chart --help cornerstones-client context --help cornerstones-client macro --help cornerstones-client events --help cornerstones-client orderflow --help ``` Common starting points: ```bash cornerstones-client fx quote --symbol EURUSD cornerstones-client fx bars --symbol XAUUSD --timeframe 1h --count 3 cornerstones-client context gold --symbol XAUUSD --timeframe 1h --count 3 cornerstones-client stocks quote --symbol AAPL cornerstones-client stocks quote --symbol 600519.SS cornerstones-client stocks normalize-symbol --symbol 600519.SH cornerstones-client chart fx --symbol XAUUSD --timeframe 15m --bars 200 --indicator ema20 --indicator rsi14 cornerstones-client events recent cornerstones-client events history cornerstones-client events receipts cornerstones-client orderflow summary --symbol XAUUSD ``` Use charts as visual evidence for multimodal agents only after checking the response metadata, artifact freshness, and account access. ## Response reading checklist Before using any market data in analysis or downstream prompts, inspect fields such as: - source labels - freshness - timestamp / as-of fields - data quality notes - availability state - account access state - source, symbol normalization, exchange, and timeframe fields when present If these fields are absent, note that absence and avoid overstating confidence. ## Operating horizon guidance Prefer 5m, 15m, hourly, daily, or higher-timeframe research workflows unless the user has separately validated a low-latency execution stack. For sub-5m trading, treat Cornerstones outputs as monitoring/context inputs, not execution guarantees. ## Do not overclaim Do not claim: - direct trade execution - guaranteed low latency - unlimited historical coverage - full Twitter/X or Reddit firehose coverage - access to Pro or Max surfaces on Free or browser trial - broker-account authority Say what the command returned, what account access applied, and what evidence fields were present. ## Example agent request pattern Use this pattern in notes and docs: ```text Ask: Give my agent a 1h XAUUSD context packet before it reasons about gold. Command: cornerstones-client context gold --symbol XAUUSD --timeframe 1h --count 3 Inspect: source labels, freshness, data quality, timestamps, and account access. Decision: proceed, downgrade confidence, choose an allowed command, or ask the user for a higher plan/key. ``` ## Common Pitfalls 1. **Skipping live help.** The command surface evolves; always run `cornerstones-client --help` and command-specific `--help` before relying on examples. 2. **Confusing keys.** A Cornerstones API key is not a broker, exchange, model-provider, or upstream data-provider key. 3. **Treating local portal config as product truth.** If `guide` or `changelog` fails because `auth status` points `portal_base_url` at a stopped local URL, report it as local configuration/runtime state and do not conclude the public guide surface is unavailable. 4. **Hiding account limits.** If a response is unavailable or requires a different account level, preserve that state in the analysis. 5. **Overstating market data.** Do not claim unlimited history, full social firehose coverage, guaranteed low latency, or execution authority unless separately verified. 6. **Treating chart artifacts as screenshots only.** Use chart artifacts as visual evidence only after inspecting freshness, source labels, artifact metadata, and account access. ## Verification Checklist - [ ] `cornerstones-client --help` ran successfully. - [ ] `cornerstones-client guide` and `cornerstones-client changelog` were inspected. - [ ] Browser trial was started for discovery, or the Cornerstones API key was stored with `auth login` and checked with `verify`. - [ ] The selected command was checked with command-specific `--help`. - [ ] The response was inspected for source labels, freshness, data quality, timestamps, account access, and symbol fields. - [ ] Any unavailable market data was reported as a constraint, not hidden.