comprehensive-research-agent

What problem does it solve? Multi-step research agents often fail silently: they ignore tool errors, cite sources never retrieved, make redundant searches, and declare research complete without validating coverage. This Skill provides structured protocols that eliminate these failure patterns in research tasks involving multiple tool calls. ## Core Features & Use Cases - Error Recovery Protocols: Mandatory acknowledgment of failed tool calls (404s, timeouts) with fallback strategies instead of silent continuation. - Source Traceability & Cross-Validation: Track which URLs were actually fetched, prohibit citing unretrieved sources, and verify key claims across at least two sources. - Structured Thinking Blocks: Require substantive reasoning traces documenting what was learned, gaps identified, and decision rationale at each step. - Use Case: When researching a topic like context engineering across 6+ tool calls (search, read_url, file writes), the agent ranks sources before reading, logs failures with recovery plans, maintains a source tracking table, and completes a pre-completion validation checklist before writing the final report. ## Quick Start Research the topic of context engineering for AI agents, validate all sources, handle any tool errors explicitly, and write a cross-validated summary report to ./output/research_summary.md.