# Execution Router

> Use when Main alone is insufficient and child, parallel, RLM, actor, or alternate-model execution is already justified — cold escalation reference.

- Skill: `ryan-brosas/execution-router` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ryan-brosas/execution-router`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ryan-brosas/execution-router/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ryan-brosas (https://skillmd.com/u/ryan-brosas)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/ryan-brosas/execution-router

---


# Execution escalation reference

Default: Main handles the task directly.

Escalate only when the task already needs it:

| Mechanism | When |
|---|---|
| Child | One bounded subsystem, reference repo, or question needs isolated context |
| Parallel workers | Independent slices with isolated writes (worktrees or non-overlapping files) |
| RLM | Context still binding after partition (Pi Fabric runner) |
| Actor | Persistent responsibility outside a one-shot task |
| Alternate model | Reasoning capability is the gap (`model-resolution`) |

Hard constraints: investigator/reviewer roles read-only; workers write only
declared scope; parallel workers do not share writable paths.

Mechanics: `fabric-native-execution`, `model-resolution`, `veda-lane`.
Evidence sourcing stays separate (`evidence-router`).