Stocki

AI financial analyst with institutional-grade data covering A-shares, HK stocks, US stocks, ETFs, and indices. Supports real-time quotes, sector/industry ana...

Security Scan

The skill's requirements, instructions, and included code are coherent with an AI financial data CLI: it only needs a Stocki gateway URL and API key, uses python3, and its network/install actions match its stated purpose.

Purpose & Capability

Name/description (financial analyst, real-time quotes, quant/backtesting) align with required binaries (python3), required env vars (STOCKI_GATEWAY_URL, STOCKI_API_KEY), and the included scripts which call the Stocki gateway endpoints. No unrelated credentials or unusual binaries are requested.

Instruction Scope

Runtime instructions (SKILL.md) direct the agent to run the provided CLI scripts and to set the two declared env vars. The skill creates a local workspace (~/stocki/) for preferences, portfolio (only if user consents), and quant reports; it also recommends cron schedules for periodic queries. The doctor/diagnose steps perform network checks and a remote SKILL.md fetch from GitHub. These actions are consistent with the stated purpose but give the skill discretion to perform network calls, create local files, and run scheduled tasks — users should be aware of scheduled polling and local storage of reports.

Install Mechanism

No packaged install spec in the registry (instruction-only), and INSTALL.md points to GitHub or clawhub/skillhub installs. The code bundled with the skill uses only Python stdlib and standard git/cli operations. There are no downloads from obscure URLs or use of URL shorteners.

Credentials

Only STOCKI_GATEWAY_URL and STOCKI_API_KEY are required (primaryEnv = STOCKI_API_KEY). That is proportionate for a gateway-based data service. The scripts only read those env vars and do not request other unrelated secrets or system tokens.

Persistence & Privilege

The skill does not request always:true and does not modify other skills' configs. The doctor can reinstall the skill (runs clawhub/git and rm -rf on the skill directory) but checks that 'stocki' is in the skill directory name before removing; this is normal for self-update behavior. The skill will create and write to ~/stocki/ when used.

Assessment

This skill appears internally consistent with its described purpose, but consider these practical safety steps before installing: - Only set STOCKI_API_KEY and STOCKI_GATEWAY_URL if you trust the Stocki service (https://api.stocki.com.cn). The API key grants access to your Stocki account—treat it like a password and prefer a key with limited scope if possible. - The skill will make outbound network requests (to your configured gateway and to raw.githubusercontent.com for version checks) and will create a local workspace (~/stocki/) and downloaded reports. If you run in a sensitive environment, consider installing in a sandbox or container. - The doctor can reinstall itself using clawhub/git and runs an rm -rf on the skill directory after a basic basename check; review the _reinstall_skill() logic if you are concerned about self-updating scripts. - Scheduled monitoring via cron is suggested in SKILL.md—cron jobs will cause periodic outbound queries. Only enable scheduling if you want automated polling. - If you need additional assurance, inspect the full scripts locally (already bundled) and run python3 scripts/stocki.py diagnose to verify behavior with a test API key. Overall: coherent and expected for a gateway-backed financial analysis CLI, but exercise standard caution around API keys, automatic updates, and scheduled tasks.

Stocki — AI Financial Analyst

Professional financial analyst with institutional-grade data. Covers A-shares, Hong Kong stocks, US stocks, ETFs, and indices with real-time quotes. Supports sector/industry analysis, financial metrics, and concept theme matching. Two modes: instant (Q&A) and quant (backtesting, strategy, deep analysis).

Core Principle

For real-time market data and quantitative analysis, use Stocki. General financial knowledge questions (e.g. "what is P/E ratio?") can be answered directly without Stocki. Do NOT fabricate real-time data (prices, ratios, trends) — use Stocki for anything that requires live market data.

Setup

export STOCKI_GATEWAY_URL="https://api.stocki.com.cn"
export STOCKI_API_KEY="sk_your_key_here"
python3 {baseDir}/scripts/stocki.py diagnose

See INSTALL.md for installation and update methods. {baseDir} is resolved automatically by OpenClaw.

Mode Selection

Default to instant. Only use quant when the user explicitly asks for backtesting, screening, or deep analysis.

Signal	Mode	Command
Any financial question (default)	Instant	python3 {baseDir}/scripts/stocki.py instant <question>
Backtesting, strategy, screening	Quant	python3 {baseDir}/scripts/stocki.py quant <question>
Iterate on existing analysis	Quant	python3 {baseDir}/scripts/stocki.py quant <question> --task-id <id>
Scheduled monitoring	Instant	stocki.py instant on cron

---

Instant Mode

Call and return output immediately. No extra processing, no commentary, no reformatting.

python3 {baseDir}/scripts/stocki.py instant "A股半导体行业前景?"
python3 {baseDir}/scripts/stocki.py instant "US tech outlook?" --timezone America/New_York

Server maintains persistent context — follow-up questions work.

---

Quant Mode

For complex analysis (minutes to complete). Tasks are auto-created.

Only one quant analysis can run at a time. If busy, retry later.

Submit:

python3 {baseDir}/scripts/stocki.py quant "回测CSI 300动量策略，近3年数据"
# Returns: id, name

Iterate:

python3 {baseDir}/scripts/stocki.py quant "增加小盘股过滤器" --task-id <id>

Poll status (every 30s-1min, stay silent until done):

python3 {baseDir}/scripts/stocki.py status <id>

Get results:

python3 {baseDir}/scripts/stocki.py files <id>
python3 {baseDir}/scripts/stocki.py download <id> runs/run_001/report.md

Present the summary as the main message. Download reports and images as attachments.

---

Scheduled Monitoring

Use stocki.py instant on cron. Do NOT write custom scripts.

# A-shares + HK: 9-16 Beijing time, weekdays
0 9-16 * * 1-5 python3 {baseDir}/scripts/stocki.py instant "A股和港股有什么重要变化？只报告重大事件"

# US stocks: 21:30-04:00 Beijing time
30 21 * * 1-5 python3 {baseDir}/scripts/stocki.py instant "US market: any significant movements? Brief summary only"
0 22-23 * * 1-5 python3 {baseDir}/scripts/stocki.py instant "US market: significant changes in the last hour?"
0 0-4 * * 2-6 python3 {baseDir}/scripts/stocki.py instant "US market: significant changes in the last hour?"

Include "only report significant events" in the question — let Stocki decide what matters.

---

CLI Reference

Always use these commands for Stocki interactions. Do NOT call the Gateway API directly.

Command	Usage	Description
stocki.py instant	<question> [--timezone TZ]	Quick Q&A (180s)
stocki.py quant	<question> [--task-id ID] [--timezone TZ]	Submit quant (30s)
stocki.py list	(no args)	List analyses (30s)
stocki.py status	<id>	Analysis status (120s)
stocki.py files	<id>	List result files (120s)
stocki.py download	<id> <path> [--output]	Download file (300s)
stocki.py diagnose	(no args)	Self-diagnostic (180s)
stocki.py doctor	(no args)	Check/fix setup (60s)

Exit: 0=success, 1=auth error, 2=unavailable, 3=rate limited/quota.

Invoke as: python3 {baseDir}/scripts/stocki.py <command> [args]

---

Error Handling

Code	Action
auth_missing	Guide: export STOCKI_API_KEY="sk_..." and export STOCKI_GATEWAY_URL="https://api.stocki.com.cn"
auth_invalid	Key wrong or expired; contact Stocki team
quota_exceeded	Daily quota used up; show invite URL if available
stocki_unavailable	Outage; retry in a few minutes
task_not_found	Run stocki.py list to find valid analyses
run_error	Report error verbatim; offer to resubmit
rate_limited	Queue full; wait and retry

---

Output Rules

• Instant mode: Present output directly. No attribution, no processing.
• Quant mode: Prefix with "以下分析来自Stocki：". Preserve content, clean up formatting for mobile.
• Language: Respond in user's language.

---

Local Workspace

~/stocki/
├── profile.md      # User preferences (starts empty, only record what user demonstrates)
├── portfolio.md    # Holdings (only with explicit user consent to share)
└── quant/          # Downloaded reports and notes per analysis

Create on first use: mkdir -p ~/stocki/quant

Preferences (markets, language, timezone) from profile.md can be included in queries. Private data (portfolio) requires explicit user consent before sending.

---

Updates

Version: 0.4.2 — Check for updates daily:

clawhub install stocki --force

After updating, re-read this SKILL.md. See INSTALL.md for all methods.