Marketplace for trusted AI agent skills

Skill Detail

Generate dated A-share research reports with TradingAgents-AShare

Submit a bounded A-share research job from OpenClaw, follow its status, and summarize the resulting analyst debate and risk report for human review.

Research & ScrapingOpenClaw
Research & Scraping OpenClaw Security Reviewed
⭐ 868 GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill generate-dated-a-share-research-reports-with-tradingagents-ashare Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
View source Documentation
At a glance
Tools required
OpenClaw, Bash, curl, Python 3; Git for source installation; TradingAgents API token and configured backend/model access
Install & setup
Install Bash, curl, Python 3 and Git. Obtain an API token from the official app Settings / API Tokens page, and configure the service and its model access as documented upstream. Install the upstream skill into the OpenClaw workspace: git clone –depth 1 https://github.com/KylinMountain/TradingAgents-AShare.git mkdir -p ~/.openclaw/workspace/skills/tradingagents-analysis cp -R TradingAgents-AShare/skills/tradingagents-analysis/. ~/.openclaw/workspace/skills/tradingagents-analysis/ Set TRADINGAGENTS_TOKEN through the runtime secret/environment configuration. Explicitly set TRADINGAGENTS_API_URL=https://api.510168.xyz for the official hosted API, or to a trusted self-hosted backend. The app.510168.xyz URL is the web UI. Do not put tokens in prompts or committed files. For self-hosting, follow the repository Docker deployment guide, set a private TA_APP_SECRET_KEY, persist backend data, and configure model access. New/modified upstream components use PolyForm Noncommercial 1.0.0; check LICENSE before commercial deployment.
Author
KylinMountain
Publisher
Open Source
Last updated
Oct 18, 2026
Quick brief

Use this OpenClaw skill when a user requests a dated research report for a Shanghai or Shenzhen A-share symbol and the agent needs to submit the analysis, follow its asynchronous job, and return a structured summary. TradingAgents-AShare combines analyst outputs with bull/bear debate and risk review. The deliverable is a research report with the service’s reported direction, confidence, available price levels, supporting arguments and risk items.

How it works

What this skill actually does

Invocation and inputs

Install the upstream tradingagents-analysis skill and its scripts directory. Supply a stock code or Chinese stock name, an explicit YYYY-MM-DD analysis date, and the requested short or medium horizon. Ask, for example: “Use TradingAgents-AShare to research 600519.SH for 2026-10-16 with the short horizon. Return the main arguments, disagreements, missing data and reported risks.” Invoke this when the agent must bring a completed report into a larger research conversation or compare a small user-specified set of symbols; the ordinary web interface remains suitable for interactive exploration.

Setup and execution

The upstream OpenClaw metadata requires Bash, curl, Python 3, TRADINGAGENTS_TOKEN and TRADINGAGENTS_API_URL. Obtain a token through the official app’s Settings / API Tokens page and provide it through the runtime environment. Explicitly select https://api.510168.xyz or an approved self-hosted backend before sending a token; https://app.510168.xyz is the web application. See the upstream README for backend/model configuration and Docker deployment.

cd ~/.openclaw/workspace/skills/tradingagents-analysis
bash scripts/analyze.sh 600519.SH 2026-10-16 short

The helper sends POST /v1/analyze, keeps the returned job_id, polls GET /v1/jobs/{job_id}, and collects GET /v1/jobs/{job_id}/result only after completed. It uses Bearer authentication and defaults to a 15-second polling interval and 600-second waiting budget. Set POLL_INTERVAL and POLL_TIMEOUT when needed, keeping polling at least 15 seconds apart. A comma-separated symbol list enables the documented batch mode; use it only for a user-specified bounded list.

Result and failure handling

Keep job IDs so an interrupted or timed-out wait can resume status checks without submitting another analysis. Treat HTTP/authentication errors, failed jobs and nonzero script exits as failures; do not interpret missing output as a successful report. A timeout does not cancel a backend job. Do not automatically repeat POST after an uncertain submission. The helper prints progress alongside formatted JSON. The API result is an envelope containing job_id, status, decision, result and finished_at; inspect the returned structure rather than assuming illustrative decision fields are always top-level. Report missing fields explicitly and preserve service-provided uncertainty, disagreement and source dates. The documented minute-scale workflow does not establish real-time market coverage.

Scope and data boundary

Submit only the extracted stock identifier, date and horizon; do not forward conversation history or local files. The backend can reuse account-held position context, so use a dedicated account without personal holdings if an analysis must be independent of stored portfolio information. Summarize generated recommendations as model output for research and human review, never as verified investment advice or a trading instruction. This skill does not place orders, import portfolios, create schedules, manage the entire application, or promise predictive accuracy. Restrict the workflow to supported A-shares and report generation.

The exact upstream skill and helper provide the invocation boundary. This differs from AShareHub’s cataloged SDK/API data retrieval: here the operator submits a multi-stage research job and collects its synthesized report. Upstream wording varies between 14 and 15 agents, so no fixed agent count is promised. The project retains Apache-2.0-derived code while new and significantly modified components use PolyForm Noncommercial 1.0.0; inspect the upstream license before commercial use.