Artificial General Intelligence

knowledge skill

Performs multi-domain problem solving by decomposing a user's goal into sub-tasks, selecting and sequencing tools or APIs, and generating executable instructions. Produces a structured plan artifact that another agent or system can run without human intervention.

Worked examples

• {"input": {"goal": "Analyze customer churn risk for Q3 and propose retention offers", "context": "Dataset: sales_transactions_2023.csv, churn_labels_2023.csv; Business rule: max discount 15%", "available_tools": ["load_csv", "train_logistic_regression", "generate_offer_codes", "export_json"]}, "output": {"plan": [{"step": 1, "tool": "load_csv", "params": {"file": "sales_transactions_2023.csv"}}, {"step": 2, "tool": "train_logistic_regression", "params": {"target": "churn", "features": ["recency", "frequency", "monetary"]}}, {"step": 3, "tool": "generate_offer_codes", "params": {"threshold": 0.7, "max_discount": 15}}, {"step": 4, "tool": "export_json", "params": {"file": "retention_offers_Q3.json"}}], "assumptions": ["Customer churn definition aligns with provided labels", "Discounts are applied at transaction level"], "risk_notes": ["Model accuracy < 80% may reduce offer effectiveness", "Discount budget capped at $50k"]}}
• {"input": {"goal": "Automate quarterly financial report generation", "context": "Templates: balance_sheet.dotx, income_statement.dotx; Data sources: gl_ledger.db, hr_system.db", "available_tools": ["query_sqlite", "render_word_template", "zip_files"]}, "output": {"plan": [{"step": 1, "tool": "query_sqlite", "params": {"db": "gl_ledger.db", "query": "SELECT * FROM trial_balance WHERE period = 'Q3'"}}, {"step": 2, "tool": "query_sqlite", "params": {"db": "hr_system.db", "query": "SELECT headcount FROM workforce WHERE quarter = 'Q3'"}}, {"step": 3, "tool": "render_word_template", "params": {"template": "balance_sheet.dotx", "data": {"trial_balance": "..."}}}, {"step": 4, "tool": "zip_files", "params": {"files": ["balance_sheet_Q3.docx", "income_statement_Q3.docx"], "output": "financial_report_Q3.zip"}}], "assumptions": ["Trial balance data is reconciled to GL", "Word templates use standard placeholders"], "risk_notes": ["Template changes may break rendering", "Data refresh delays could impact timeliness"]}}

Input

• goal: string: user's high-level objective
• context: string: domain-specific background or constraints
• available_tools: array: list of tool/function names the agent may invoke

Output

• plan: array: sequence of steps with tool invocations and parameters
• assumptions: array: key assumptions made during planning
• risk_notes: array: potential failure points and mitigations

Details

• Skill type: knowledge skill
• Safety level: safe_public_research
• Version: 1.0.0