What problem does it solve? Researchers lose track of what has already been tried, how experiments relate to each other, and which past results have been retracted. This Skill answers "what's been done in this area" questions by searching and browsing the Chronicle experiment platform read-only, so new work builds on verified prior results instead of repeating them. ## Core Features & Use Cases - Semantic search over internal history: Finds past experiments and attached reports (hypothesis, takeaways, research) by topic using chronicle.search, with optional author and time-window filters. - Structural browsing and lineage walks: Lists experiments by lifecycle state (open, committed, concluded) and renders the parent/child DAG for any experiment via chronicle.get_lineage, including pending tentative fork links. - Retraction provenance: Flags retracted ancestors in a lineage so users know when a result's premises have been invalidated. - Use Case: Before proposing a new experiment, ask "what have we tried on learning-rate schedules for this model?" and get a compact list of relevant past experiments, their states, and any retracted upstream work. ## Quick Start Ask the agent to show the lineage and any retracted ancestors of experiment exp_123, or to search experiment history for a topic like "what experiments exist about gradient clipping".
