# librarian-agent-instantiation

> How to spawn and task the Librarian Agent for media ingestion

- Author: NOVA
- Repository: NOVA-Openclaw/clawd-workspace
- Version: 20260207131035
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-07
- Source: https://github.com/NOVA-Openclaw/clawd-workspace
- Web: https://mule.run/skillshub/@@NOVA-Openclaw/clawd-workspace~librarian-agent-instantiation:20260207131035

---

---
name: librarian-agent-instantiation
description: How to spawn and task the Librarian Agent for media ingestion
---

# Librarian Agent Instantiation

## Tools

- `sessions_spawn`
- `psql`
- `read`
- `web_fetch`
- `exec`

## Procedure

1. GATHER INPUT: Get URL, file path, or queue item ID to process
2. FETCH AGENT CONFIG: SELECT * FROM agents WHERE name = 'librarian-agent'
3. CHECK QUEUE (optional): SELECT * FROM v_media_queue_pending LIMIT 5
4. CHECK DUPLICATES: SELECT id, title FROM media_consumed WHERE url = '{input_url}'
5. IF DUPLICATE: Return existing record, skip spawning
6. BUILD CONTEXT: Read SOUL.md and WORKFLOW.md from seed_context.files
7. SPAWN AGENT: sessions_spawn(task="{context_template}\n\nPending Queue:\n{queue_items}\n\nTask: Ingest this media: {input}", model="claude-sonnet-4-20250514", label="librarian:{media_slug}", cleanup="delete")
8. MONITOR: Use sessions_list to check progress
9. VERIFY: Confirm media_consumed record was created
10. REPORT: Return media_consumed.id and summary to caller

## Notes

Use for individual media items or batch queue processing. Agent handles deduplication internally.