What problem does it solve? Authoring agent components for Claude Code, Cursor, Codex, and Cowork from memory produces spec drift: skills that fail Cowork upload, rules Cursor silently ignores, and MCP configs Codex never loads. This Skill pins every authoring decision to archived primary sources so forged components actually load where they are supposed to. ## Core Features & Use Cases - Seven-Stage Forge Pipeline: Topic, research, distillation, references, guides, skill file, and registration, with each component's domain research archived and cited. - Cross-Harness Authoring Guides: Per-type, per-harness field tables, working examples, and gotchas covering all 20 component-harness combinations, plus a support matrix for format decisions. - Validation and Packaging Scripts: Deterministic Python scripts validate any component per harness and package skills and plugins for Cowork upload. - Use Case: Ask to forge a new Stinger for Postgres migration safety across all four harnesses, and the Skill walks the full pipeline, validates the output, and registers the Bee-Stinger pair with the beekeeper roster. ## Quick Start Ask the orchestrator to forge a new skill for API contract testing across Claude Code, Cursor, Codex, and Cowork, then register it with the beekeeper.