agent-research-aggregator

To aggregate scattered AI agent logs into structured paper inputs, this skill scans common agent caches like .claude and .cursor, extracts experiments and numeric results via batch LLM-assisted processing, and formats them into deterministic idea.md and experimental_log.md files.

Extracting experiment data from fragmented coding agent histories is done by performing a read-only discovery scan across workspace caches, consolidating the findings into a single synthesis JSON, and applying schema validation to generate reviewable paper drafting inputs.

Yes, you can consolidate logs from multiple agent caches simultaneously. The skill performs a read-only scan across .claude, .cursor, .antigravity, and .openclaw directories to locate scattered experiment files and metrics for unified extraction.

The log aggregation process does not modify your original agent cache files. It performs strictly read-only discovery and extraction, ensuring your source experiment histories remain untouched while generating new structured PaperOrchestra inputs.

You validate data quality when converting agent logs by utilizing the skill's built-in schema validation, truncation safeguards, and confidence tagging features, which collectively generate an optional aggregation report highlighting file counts, warnings, and overall confidence.

Limitations when synthesizing unstructured chat histories include potential truncation safeguards activating on excessively large logs and variable confidence tagging, requiring user review of the generated idea.md and experimental_log.md to ensure accurate paper drafting.