Scenario Analyzer

ニュースヘッドラインを入力として18ヶ月シナリオを分析するスキル。 scenario-analystエージェントで主分析を実行し、 strategy-reviewerエージェントでセカンドオピニオンを取得。 1次・2次・3次影響、推奨銘柄、レビューを含む包括的レポートを日本語で生成。 使用例: /scenario-analyzer "Fed raises rates by 50bp" トリガー: ニュース分析、シナリオ分析、18ヶ月展望、中長期投資戦略

This skill appears coherent for its stated purpose, but check these operational points before installing: - Dependents: The skill calls two subagents (scenario-analyst and strategy-reviewer). Verify those subagents exist and inspect their implementations (they could request env vars or network access that this orchestration doc doesn't show). - Network use: The workflow requires WebSearch (past 2 weeks) and invoking the platform Task tool; expect outbound network/search queries for each run. If you have privacy concerns about headlines or proprietary text being sent to external search, avoid using confidential inputs. - File writes: It will create a reports/ directory and save markdown reports there. Make sure writing files to the agent workspace is permitted and that no sensitive data will be stored inadvertently. - Financial advice: The skill selects and recommends US tickers; if you are concerned about regulatory/ethical implications, treat outputs as informational, not professional investment advice. - Autonomy: The skill is user-invocable and not forced-always; if you prefer to limit autonomous actions, restrict model invocation or require explicit consent before running subagents or WebSearch. If you want a deeper safety check, provide the implementations (code) for the scenario-analyst and strategy-reviewer agents so they can be reviewed for unexpected credential access, hidden endpoints, or unsafe shell/network operations.

Type: OpenClaw Skill Name: scenario-analyzer Version: 0.1.0 The OpenClaw AgentSkills skill bundle is designed to analyze news headlines and generate investment scenarios. All files, including the `SKILL.md` instructions, are clearly aligned with this stated purpose. The agent is instructed to perform `WebSearch` for news gathering and to write a report to a dedicated `reports/` directory, explicitly prohibiting saving to the project root. These capabilities are necessary for the skill's function and are not indicative of malicious intent or high-risk behavior beyond the stated purpose. No evidence of data exfiltration, malicious execution, persistence, or prompt injection to subvert the agent's core mission was found.