What problem does it solve? Finding grounded evidence across large collections of documents (interviews, PDFs, transcripts, notes) is slow and error-prone, and AI answers often lack traceable sources. This Skill systematically searches a raw corpus using maps and a term dictionary, then writes every finding to a citable evidence packet file. ## Core Features & Use Cases - Guided corpus navigation: Reads a term dictionary and structural maps before searching, expanding queries with synonyms, grammatical forms, and colloquial-to-technical term bridging. - Bounded iterative search: Runs up to 5 grep rounds with early-stop conditions (sufficient evidence, diminishing returns, token budget) and decomposes complex comparative queries into sub-queries. - Structured evidence packets: Writes findings to agent_reports/ files with YAML frontmatter tracking sources, confidence levels, navigation paths, and search termination reasons. - Use Case: A researcher asks "What causes geomagnetic reversals?" across a folder of earth science notes; the Skill consults the dictionary, searches the corpus, and returns a file path to an evidence packet with quoted excerpts and confidence ratings. ## Quick Start Ask the agent to search the workspace corpus for evidence on your research question and return the path to the written evidence packet.
