"Use backend-platform for Unstract's Django/DRF APIs, hosted MCP
Scanned 9/9/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill backend-platform --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Backend Platform?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-backend-platform-arex-skill)More formats (shields.io, HTML) on the badges page.
---
name: "backend-platform"
description: "Use backend-platform for Unstract's Django/DRF APIs, hosted MCP
servers, route wiring, auth, and backend configuration."
disable-model-invocation: true
metadata:
disco-role: operating
license: AGPL 3.0
---
# Backend Platform
Use this sub-skill when the task is about the Django backend itself: URL routing, auth, API deployment, pipeline APIs, file management, platform API keys, or the hosted MCP server.
## Owns
- Backend URL composition in `backend/backend/*.py`.
- API families such as `account_v2`, `api_v2`, `pipeline_v2`, `file_management`, `platform_api`, and related route wiring.
- The hosted MCP server, tool registry, tool authorization, spend guard, and redaction tests.
- Backend settings, environment-variable contracts, and developer / test startup checks.
## Excludes
- Full-stack deployment and container bootstrap — use `platform-deployment`.
- Worker queue routing and Celery / PG-queue operations — use `workers`.
- Frontend routing or runtime config — use `frontend`.
- Shared tool / SDK / registry authoring — use `sdk-and-tools`.
- Test-group manifests and e2e runtime selection — use `testing-rig`.
## Start Here
Read `references/routes-and-mcp.md` first for route families, MCP URLs, auth behavior, and the tool catalog.
Read `references/configuration.md` for the backend env var contract, startup command shape, and safe inspection environment.
Read `references/troubleshooting.md` when the issue is a failed import, a missing backend env var, a route mismatch, or an MCP authorization problem.
For a safe backend smoke check, use the root checker from this skill tree:
```bash
python ../../scripts/check_unstract_packages.py --backend
```
## Shared References
- `references/routes-and-mcp.md` — route families, hosted MCP topology, tool catalog, and auth behavior.
- `references/configuration.md` — settings contract, env vars, and startup commands.
- `references/troubleshooting.md` — backend import, routing, auth, and hosted-MCP failures.
- `../../references/service-map.md` — repo-wide service ownership map.
- `../../references/installation-and-env.md` — install and env matrix for backend and adjacent services.
- `../../scripts/check_unstract_packages.py` — shared package / backend smoke checker.
## Common Task Routing
| User request | Read next |
| --- | --- |
| "Why does this route 404 or 403?" | `references/routes-and-mcp.md` |
| "How do the MCP servers differ?" | `references/routes-and-mcp.md` |
| "What env vars does the backend need?" | `references/configuration.md` |
| "Why did the backend import fail?" | `references/troubleshooting.md` |
| "What should the API deployment or pipeline route look like?" | `references/routes-and-mcp.md` |
## Safety Boundaries
- Do not assume live PostgreSQL, Redis, or RabbitMQ are available unless the user asks for runtime validation.
- Do not suggest direct access to credential-bearing endpoints that the hosted MCP server intentionally excludes.
- Do not rely on direct view calls for middleware-authenticated behavior; always use the real URL path when checking auth semantics.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!