---
title: "starlight-department-lab"
description: "Practice human judgment, agentic systems and scientific inference with portable mission records; design departments and test workflow evidence."
canonical: https://agentpluginsdirectory.com/plugins/starlight-department-lab
last-updated: 2026-09-13
---

# starlight-department-lab
Practice human judgment, agentic systems and scientific inference with portable mission records; design departments and test workflow evidence.
- Slug: starlight-department-lab
- Publisher: Starlight Intelligence Systems
- Repository: https://github.com/frankxai/starlight-academy-plugin
- Manifest: plugins/starlight-department-lab/plugin.json
- Version: 0.1.0+codex.20260907044049
- License: Apache-2.0
- Category (editorial): other
- Skills: 5 (design-ai-department, learn-starlight-mission, practice-agent-evaluation, test-workflow-evidence, verify-ai-department-package)
- MCP servers: 1 (starlight-department-lab)
- Stars: 0
- Repository created: 2026-09-12
- Repository last pushed: 2026-09-12
- Publisher type: User
- Listing: https://agentpluginsdirectory.com/plugins/starlight-department-lab
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What starlight-department-lab does, in the publisher's words

Five skills and eight local, read-only MCP tools for four practice missions. Preserve an initial attempt, assistance, artifact, critique, revision and transfer. The server uses bundled public examples. It makes no network calls or external actions. Practice is unsigned self-study, not certification or permission to act.

Website: https://starlightintelligence.academy/academy/studio

This plugin and its bundled learning material are Apache-2.0; see LICENSE. Other repository content retains its own license.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/frankxai/starlight-academy-plugin/HEAD/plugins/starlight-department-lab/README.md

## Skills

- design-ai-department: Triage and design one accountable AI department from a recurring workflow, including when not to automate, the smallest viable topology, human authority, evidence, evaluations, recovery, and portable package projections. Use when asked to design an AI team, agent department, governed workflow, mult…
- learn-starlight-mission: Complete or facilitate a versioned Starlight Academy mission for a human or sponsored agent, producing the required artifact and an honest evidence packet without claiming certification or runtime authority.
- practice-agent-evaluation: Guides humans and agents through synthetic workflow evaluation, preserving an unaided attempt, checking sources and authority, revising judgments, and testing transfer. Use for agent outcomes, approval scope, or retrieved-source instructions.
- test-workflow-evidence: Build and revise deterministic acceptance checks for one synthetic agent workflow using public fixtures, false acceptances, false rejections and explicit unknown outcomes. Use after evaluation practice when a learner wants executable tests, an inspectable test suite or a regression case.
- verify-ai-department-package: Independently audit an AI department packet or installable package for workflow fit, authority, topology simplicity, producer-verifier separation, evidence, evaluations, recovery, security, portability, and honest release claims. Use before publishing, installing, admitting, deploying, or materiall…

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.

## MCP servers

- starlight-department-lab: transport: stdio; command: node ${PLUGIN_ROOT}/mcp/server.mjs

Read from the plugin's own mcp.json. Environment variable names only, never values.