Knuckles-Team Research Scanner Workflow
CONCEPT:DEV-001
Discover research papers using ScholarX and bulk ingest them into the Graph-OS Knowledge Graph
Steps
Step 0: collect-research-request [skill: user-interaction]
Agent: scanner-agent
Tools: rep_rm_workspace, rep_rm_git
Get the user's research focus query, lookback days, specific target scientific sources (e.g., arxiv, biorxiv), and maximum results.
Step 1: search-papers [skill: scholarx-mcp]
Agent: builder-agent
Tools: rep_rm_projects
Search for the research papers based on the search parameters using the scholarx_search action with action='search'.
Step 2: download-papers [skill: scholarx-mcp]
Agent: validator-agent
Tools: rep_rm_projects, gl_pipelines
Submit bulk download jobs for the discovered paper IDs via scholarx_storage with action='bulk_download'. Check the download queue status via action='status'.
Step 3: Graph Os
Agent: publisher-agent
Tools: rep_rm_git, gl_merge_requests
Ingest the newly downloaded research PDFs into the Knowledge Graph via the ingest action mcp_graph-os_graph_ingest with action='ingest'.
Step 4: present-research-summary [skill: user-interaction]
Agent: scanner-agent
Tools: rep_rm_workspace, rep_rm_git
Display the completed research summary and details of the ingested papers to the user.
Step 5: KG Persistence [depends_on: Step 0, Step 4]
Agent: publisher-agent
Tools: graph_write
Persist workflow results as nodes and edges in the Knowledge Graph. Create appropriate typed nodes with metadata and link to existing domain entities.
Output
- Research Scanner results persisted in KG
- Structured report (MD/PDF)
- Audit trail with timestamps and agent attributions
Execution
Run this workflow as a dependency-ordered DAG. Steps with no unmet depends_on run in parallel; dependents run after their prerequisites complete.
- Run first (in parallel): Step 0 — collect-research-request; Step 1 — search-papers; Step 2 — download-papers; Step 3 — Graph Os; Step 4 — present-research-summary
- After level 0: Step 5 — KG Persistence
Execution: If graph-os is reachable, offload the whole DAG via graph_orchestrate action=execute_workflow (or the kg-delegate skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet depends_on in parallel, then their dependents.