What problem does it solve? Teams extending an agent harness system often write bloated, overlapping, or mis-scoped skills and subagents that waste context budget and cause false triggers. This Skill provides the methodology, decision rules, and scaffolding templates to author new skills and subagents correctly. ## Core Features & Use Cases - Skill vs. Subagent Decision Framework: Provides comparison tables and a quick decision matrix to decide whether a capability should be a context-injected skill or an isolated-execution subagent. - Context Budget Discipline: Enforces a three-tier loading mechanism (resident metadata, triggered body under 500 lines, on-demand references) with concrete splitting strategies. - Scaffolding & Quality Gates: Ships scaffold templates, design pattern references, and an automated check script that validates structure, reference files, and required sections. - Use Case: When a user asks "should this be a skill or a subagent" or wants to slim a SKILL.md that has grown past 500 lines, this Skill routes the decision, extracts complete sections into references/ files, and updates navigation pointers. ## Quick Start Ask the agent to scaffold a new harness skill named "deploy-checker" following the harness-authoring conventions and check it for overlap with existing skills.
