What problem does it solve? Over time an AI agent accumulates hundreds of sessions with cryptic auto-generated titles, making it impossible to find past decisions, and cluttering the library with stale or duplicate conversations. This Skill manages the session library conversationally: it finds sessions by topic, summarizes what was decided, renames them meaningfully, and safely archives or deletes the ones you no longer need. ## Core Features & Use Cases - Topic-Based Discovery: Uses full-text search (session_search) over the message store plus metadata filters (hermes sessions list) to locate sessions by content, age, source, cost, or workspace. - Safe Cleanup Workflow: Presents a plan table before any mutation, prefers reversible archive over delete, and requires --dry-run plus explicit confirmation before destructive prune or delete commands. - Meaningful Renaming & Export: Renames sessions with short, unique, prefix-friendly titles for /resume, and exports valuable sessions to Markdown before deletion. - Parallel Workstreams: Splits work into one delegated task per ticket via delegate_task, with each subagent transcript remaining searchable later. - Use Case: Ask "find my sessions about Q3 pricing, keep the useful ones, and clean up the duplicates" and receive a summarized plan of renames, archives, and proposed deletions to approve before anything changes. ## Quick Start Ask the agent to find all sessions about a specific project, summarize what was decided in each, and propose which stale duplicates to archive.
