Skip to content
Back to skills

Os Delegation Router V1.0

ASecurity

Route tasks to native OS, ACS, or TDE

  • 14 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 3, 2026
ai-agents

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 3, 2026

npx -y skills add CorvinLabs/CorvinOS --skill os_delegation_router_v1.0 --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Os Delegation Router V1.0?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Os Delegation Router V1.0
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/corvinlabs-os-delegation-router-v1-0/badge)](https://www.skillsdirectory.com/skills/corvinlabs-os-delegation-router-v1-0)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: os.delegation_router
version: "1.0.0"
description: Route tasks to native OS, ACS, or TDE
---

# OS-Skill: Delegation Router v1.0

You are the delegation router skill. Your job: decide where to route a task.

## Phase 0 — Intake

Validate input against input_schema:
- task_shape: one of [small_code, big_data, prose, structured]
- context_size: integer 0-2000000
- tenant_id: alphanumeric + underscore/dash

If any field missing or invalid type, report error and exit.

## Phase 3 — Plan + Decision

**Heuristic (MVP):**
- If task_shape == "big_data": route to ACS (cost-efficient for large data)
- Else: route to native (lower latency)

Set confidence based on certainty:
- big_data heuristic: confidence 0.75 (data efficiency is reliable)
- native fallback: confidence 0.80 (native is always an option)

Reasoning: brief one-sentence explanation

## Phases 1, 2, 4-9

Skipped in MVP (no context loading, no clarification, no full logic flow).

## Phase 10 — Output

Return JSON:
```json
{
  "decision": "acs" or "native" or "tde",
  "confidence": float 0.0-1.0,
  "reasoning": "string explanation"
}
```

Validate against output_schema. If invalid, error.

## Phase 11 — Learning

Skipped in MVP (done in Phase 2).

---

## No Feedback Loop in MVP

Feedback ingestion and optimization added in Phase 2.

Files in this skill

  • SKILL.md1.3 KB
  • manifest.yaml2 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.