ai-coding-agents-runtime-core
AI Coding Agents Runtime Core
Use this skill to design or review the two registries every coding-agent CLI ships: the tool runtime (tool contracts, pool assembly, deferred loading, ToolSearch, execution pipeline, remote rendering, subagent dispatch) and the slash-command runtime (typed command kinds, source-aware discovery, lazy loading, forked commands, remote-safe dispatch).
Both are registries with sources, precedence, filtering before exposure, refresh after topology change, and a single execution path. Design them with one contract vocabulary so permission, telemetry, and rendering semantics do not diverge.
This skill owns tool and command architecture. It does not own approval policy or sandboxing (../ai-coding-agents-safety-envelope/SKILL.md), plugin packaging (../ai-coding-agents-plugins/SKILL.md), session persistence (../ai-coding-agents-state/SKILL.md), MCP server design (../agents-mcp/SKILL.md), or broader architecture (../ai-coding-agents/SKILL.md).