curating-memories
About
This Claude Skill provides guidance for maintaining memory quality through curation, including updating outdated content, marking obsolete information, and linking related knowledge. Use it when memories need modification, when new information supersedes old data, or when building knowledge graph connections between related concepts.
Quick Install
Claude Code
Recommended/plugin add https://github.com/majiayu000/claude-skill-registrygit clone https://github.com/majiayu000/claude-skill-registry.git ~/.claude/skills/curating-memoriesCopy and paste this command in Claude Code to install this skill
Documentation
Curating Memories
Active curation keeps the knowledge base accurate and connected. Outdated memories pollute search results and reduce effectiveness.
When to Update a Memory
Use update_memory when:
- Information needs correction or clarification
- Importance level changes (more/less relevant than thought)
- Content needs refinement
- Links to projects/artifacts/documents change
execute_forgetful_tool("update_memory", {
"memory_id": <id>,
"content": "Updated content...",
"importance": 8
})
Only specified fields are changed (PATCH semantics).
When to Mark Obsolete
Use mark_memory_obsolete when:
- Memory is outdated or contradicted by newer information
- Decision has been reversed or superseded
- Referenced code/feature no longer exists
- Memory was created in error
execute_forgetful_tool("mark_memory_obsolete", {
"memory_id": <id>,
"reason": "Superseded by new architecture decision",
"superseded_by": <new_memory_id> // optional
})
Obsolete memories are soft-deleted (preserved for audit, hidden from queries).
When to Link Memories
Use link_memories when:
- Concepts are related but not caught by auto-linking
- Building explicit knowledge graph structure
- Connecting decisions to their implementations
- Relating patterns across projects
execute_forgetful_tool("link_memories", {
"memory_id": <source_id>,
"related_ids": [<target_id_1>, <target_id_2>]
})
Links are bidirectional (A↔B created automatically).
Curation Workflow
When creating new memories, check impact on existing knowledge:
Step 1: Query Related Memories
execute_forgetful_tool("query_memory", {
"query": "<topic of new memory>",
"query_context": "Checking for memories that may need curation",
"k": 5
})
Step 2: Analyze Each Result
For each existing memory, determine action:
| Situation | Action |
|---|---|
| Existing memory is still accurate | Link to it |
| Existing memory has minor gaps | Update it |
| Existing memory is now wrong | Mark obsolete, create new |
| Existing memory is partially valid | Create new, link both |
Step 3: Execute Curation Plan
Present plan to user before executing:
Curation plan:
- Create: "New authentication approach" (importance: 8)
- Mark obsolete: #42 "Old auth pattern" (superseded)
- Link: New memory ↔ #38 "Security requirements"
Proceed? (y/n)
Step 4: Execute and Report
After user confirms:
- Create new memory
- Mark obsolete memories
- Create links
- Report results with all changes made
Signs of Poor Curation
Watch for these indicators:
- Multiple similar memories on same topic (deduplicate)
- Memories referencing deleted code (mark obsolete)
- Contradictory memories (resolve conflict)
- Low-importance memories (importance < 6) accumulating
- Orphaned memories with no links (consider linking or removing)
Auto-Linking
Forgetful auto-links semantically similar memories (similarity >= 0.7) during creation. Manual linking is for:
- Explicit relationships auto-linking missed
- Cross-project connections
- Non-obvious conceptual links
Check similar_memories in create response to see what was auto-linked.
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
creating-opencode-plugins
MetaThis skill provides the structure and API specifications for creating OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It offers implementation patterns for JavaScript/TypeScript modules that intercept and extend the AI assistant's lifecycle. Use it when you need to build event-driven plugins for monitoring, custom handling, or extending OpenCode's capabilities.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
evaluating-llms-harness
TestingThis Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.
