session-skill-mining

What problem does it solve?

It helps identify and crystallize reusable “how-to” procedures and durable project/user knowledge from prior AI coding sessions, so future agent runs don’t waste time rediscovering the same decisions, constraints, and workflows.

Core Features & Use Cases

• Skill discovery for reusable procedures: Extracts ordered, role-shaped workflows (e.g., routing, delegation, scheduling, verification rituals) that can be turned into new SKILL.md files.
• Memory extraction for durable facts: Captures long-lived project or user constraints/preferences (e.g., deployment quirks, ownership boundaries, batching habits) as memory-project or memory-user.
• Strategy-shaped optimization coverage: Converts strategy-like signals (cost/routing/scheduling/role-boundary friction) into skills or memories rather than emitting a separate strategy result.
• Auto-optimizer compatibility: When invoked by the auto-optimization scheduler, produces structured JSON suggestions suitable for parsing/persistence.

Quick Start

Use session-skill-mining to scan a specific time range of your Codex or Claude Code activity and produce high-signal skill and memory candidates in a single response.