resilient-research-workflow

Resilient Research-to-Spreadsheet Workflow

Use this skill when you must gather web-based information AND produce verified Excel workbooks, with particular emphasis on handling tool failures gracefully and proving final outputs against explicit criteria.

This skill combines:

• Phase A: Resilient web research with automatic failure detection and shell_agent delegation
• Phase B: Research data validation before spreadsheet integration
• Phase C: Anchored spreadsheet proof gate for workbook creation/editing
• Cross-phase gates: Explicit handoff criteria between research and spreadsheet phases

This is an end-to-end workflow skill. It requires:

• autonomous retry logic for web research (via shell_agent when needed)
• direct Python verification of research outputs
• deterministic spreadsheet proof methodology before finalizing

When to use

Use this skill when ANY of the following are true:

Research-phase triggers:

• You need current web-based data (prices, market info, competitor analysis, etc.)
• Prior search_web calls have failed or returned inconsistent results
• The task requires multi-step data collection from multiple sources
• You need resilient data gathering that can adapt to tool instability

Spreadsheet-phase triggers:

• You must create or modify an Excel workbook based on research data
• The user requires exact sheet names, columns, counts, or samples
• Prior summaries are inconsistent or not trustworthy
• Exact post-edit workbook state matters
• You must prove that mandatory criteria were satisfied

Both phases:

• A delegated agent may perform parts of the work
• Tool instability has been observed in the task context
• The task has failed previously due to tool errors

Core rules

Rule 1: Detect and escalate research failures

When search_web returns errors like:

• "unknown error"
• Empty or incomplete results
• Consistent failures across multiple queries (>2 failures)

Immediately delegate to shell_agent rather than retrying the same failing tool.