knowledge-import-pipeline

name	knowledge-import-pipeline
description	Import external documents, repositories, notes, transcripts, or datasets into the appropriate Prometheus memory, project context, skill resource, or temporary artifact. Use for deliberate knowledge ingestion with provenance; do not use for merely reading or summarizing a file.

Knowledge Import Pipeline

Use this skill when bringing external information into Prometheus.

Current State

Status: mostly usable manually; needs a dedicated import wizard/tool for 110%.

Current support:

• Memory tools: memory_search, memory_read_record, memory_index_refresh, memory_write, write_note
• Evidence index source types include chat/session/task/proposal/cron/team/project/memory roots.
• Operational records include decisions, preferences, project facts, task outcomes, workflow rules, entity facts, and conversation summaries.
• Memory index files live under workspace/audit/_index/memory/.
• Project context and team context references can store scoped knowledge.
• Extension runtime supports registerMemorySource(...), but source use must be verified before relying on it.

Procedure

1. Identify import target: durable memory, project context, team context, skill, artifact, or temporary note.
2. Classify content:
   • entity/business fact -> entity or BUSINESS/project context
   • user preference -> USER memory
   • workflow/procedure -> skill/resource, not general memory
   • one-off evidence -> artifact/note
3. Chunk or summarize source material with source paths/URLs.
4. Write only durable facts to memory.
5. Run memory_index_refresh after meaningful imports.
6. Verify with memory_search and memory_read_record.

Guardrails

• Do not dump raw docs into durable memory.
• Preserve citations/source paths.
• Keep temporary research separate from stable facts.