What problem does it solve?
Investigating an entity (a username, file, symbol, ADR, URL, or concept) usually means manually querying many sources and stitching results together. This Skill automates recursive, multi-source entity expansion and produces a provenance-tracked graph dossier instead of a one-off answer.
Core Features & Use Cases
- Parallel fan-out investigation: Issues all source queries (WebSearch, WebFetch, Grep, Glob, memory search, embeddings search) in a single round, then recurses breadth-first up to a configurable depth and breadth.
- Graph output with provenance: Aggregates results into a nodes/edges graph where every node and edge records its source and confidence, rendered as both a Markdown report with a mermaid diagram and a machine-readable JSON file.
- Budget and dedup discipline: Supports cost budgets with explicit truncation marking, plus exact and embedding-similarity-based entity deduplication.
- Use Case: Run it on a GitHub username to map owned repositories, related contributors, and linked concepts, or on an ADR identifier to trace every file and decision referencing it.
Quick Start
Ask the agent to build a dossier on the seed 'ruvnet' with a maximum depth of 2 and a budget of one dollar.
