What problem does it solve? Hermes deployments accumulate session data, skills, cron jobs, and costs over time, but there is no built-in way to systematically audit that history and turn it into concrete optimization recommendations. This Skill gathers live evidence from the session database and skill library, then produces either a direct review or a structured analyst prompt for an external model. ## Core Features & Use Cases - Session store analysis: Queries ~/.hermes/state.db for usage statistics by source, model, cost, and recurring task patterns. - Runaway session diagnostics: Diagnoses a specific session that hit tool-call guardrails, ignored stop instructions, or replayed stale tool calls after compaction. - Analyst prompt generation: Composes a self-contained, structured prompt that an external model can use to audit the deployment, saved to a known file location. - Evidence-first validation: Separates observed symptoms from confirmed root causes and verifies that proposed configuration changes actually exist before recommending them. - Use Case: After noticing rising token costs, ask for a system audit; the Skill aggregates session statistics, maps the skill library, and writes a prioritized findings report with evidence and expected impact. ## Quick Start Ask the assistant to audit your Hermes setup by analyzing session history, skill usage, and token costs, then produce a prioritized list of optimization recommendations.
