agent-research-aggregator

To turn scattered AI agent logs into research paper inputs, this skill scans cache directories like .claude and .cursor to discover experiment records, extracts structured data via LLM synthesis, and formats them into idea.md and experimental_log.md files for paper writing.

Yes, you can automatically extract experiment records from .claude, .cursor, .antigravity, or .openclaw cache folders. The skill performs a read-only scan to find relevant logs and uses an LLM-assisted pass to build a consistent JSON representation of hypotheses, methods, and results.

The best way to prepare PaperOrchestra workspace inputs from past agent runs is to run this skill with your agent-log root directory. It consolidates extracted experiments into a synthesis.json and deterministically generates the required workspace/inputs/idea.md and experimental_log.md files.

No, the agent log aggregation process does not overwrite existing files. It includes a prevention mechanism that skips entirely when workspace/inputs/idea.md and workspace/inputs/experimental_log.md already exist and are non-empty, avoiding unnecessary work.

Yes, you need Python to format experimental logs for paper writing, as the skill depends on Python, json, and pathlib to discover files, extract experiment records, and synthesize them into structured markdown inputs.

Agent log extraction requires a project selection step because the skill enforces a project-selection gate before initiating LLM extraction. This ensures the aggregation process targets the correct experimental context before generating your workspace inputs.