knowledge-tidying

Knowledge Tidying Skill

Delegates to the knowledge-tidier agent (haiku). See registry.yaml.

When This Skill Activates

• Always after research-synthesis completes (required step)
• After major investigation concludes with valuable findings
• When user says "save this", "persist findings", "remember this"
• When valuable insights should be preserved for future sessions
• Consolidating or organizing existing Nexus knowledge (T3)

Agent Invocation

Use the Agent tool to invoke knowledge-tidier:

## Relay: knowledge-tidier

**Task**: [what needs to be done]
**Bead**: [ID] or 'none'

### Input Artifacts
- Files: [relevant files]

### Deliverable
Organized knowledge entries in T3

### Quality Criteria
- [ ] Knowledge stored in Nexus T3 via store_put tool
- [ ] No contradictions with existing knowledge
- [ ] Tags are meaningful for future retrieval

For full relay structure and optional fields, see RELAY_TEMPLATE.md.

Nexus Storage Standards

Store to T3 knowledge (store_put tool):

• store_put tool: content="# content", collection="knowledge", title="research-topic", tags="research"
• store_put tool: content="# content", collection="knowledge", title="decision-component-name", tags="decision,architecture"
• store_put tool: content="# content", collection="knowledge", title="pattern-name", tags="pattern"

Title conventions:

• research-{topic} - Research findings
• debug-{component}-{issue-type} - Debugging insights
• architecture-{project}-{component} - Architecture documentation
• decision-{component}-{decision-name} - Architectural decisions
• pattern-{pattern-name} - Reusable patterns

Verify storage:

• mcp__plugin_nx_nexus__query(question="topic", content_type="knowledge") — confirm searchable via catalog-scoped query
• mcp__plugin_nx_nexus__store_list(collection="knowledge" — list all knowledge entries

Contradiction Handling

If contradictions found with existing knowledge:

1. Search: mcp__plugin_nx_nexus__search(query="topic", corpus="knowledge" to find related entries
2. Identify which is more current/accurate
3. Replace the stale entry by re-storing with the corrected content

Agent-Specific PRODUCE

• Consolidated Documents: Store in nx T3 via store_put tool: content="# {topic}\n{consolidated-content}", collection="knowledge", title="consolidation-{date}-{scope}", tags="consolidation,tidier"
• Archive Actions: Moved documents go to --collection knowledge__archive --title "{old-title}-archived-{date}", logged in nx T2 memory as --project {project} --title archive-log.md
• Contradiction Resolutions: Updated directly in source nx T3 documents
• Review Artifacts: Use T1 scratch to track review round findings:
  • scratch tool: action="put", content="# Review Round {N}: {N} issues found\n{issue-list}", tags="review,round-{N}"
  • scratch_manage tool: action="promote", entry_id="", project="{project}", title="review-round-{N}.md"

Success Criteria

• Knowledge stored in Nexus T3 via store_put tool
• No contradictions with existing knowledge
• Knowledge is searchable (verify with search tool)
• Tags are meaningful for future retrieval
• Title follows naming convention