What problem does it solve? Evaluating external graph-analysis tools like Graphify or InfraNodus normally risks corrupting canonical workspace data. This Skill runs those tools against temporary isolated copies of the workspace, capturing outputs safely so you can assess tag extraction and graph exploration quality without touching real knowledge data. ## Core Features & Use Cases - Isolated spike execution: Copies only canonical files (cards/, refs/, connections.json, domains.yaml) to a temp directory before running the external tool, then cleans up automatically. - Structured result capture: Writes stdout, stderr, output files, duration, and success status to log/spike-results/{tool}-{timestamp}.json for later review. - Evaluation frameworks: Provides per-spike criteria for Graphify-style tag extraction (precision, recall, novelty) and InfraNodus-style graph exploration (structural insight, UI applicability). - Use Case: Run construct spike run graphify --workspace test-ws/my-construct to see whether extracted candidate tags are novel compared to existing search-seeds, then document findings as design input. ## Quick Start Ask the agent to run a graphify spike against your test workspace and review the captured results file in log/spike-results.
