NekoCore-OS search-archive
Search long-term memories that have moved to the archive tier (> 90 days old, low-activity). Returns ranked results by semantic relevance.
install
source · Clone the upstream repo
git clone https://github.com/voardwalker-code/NekoCore-OS
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/voardwalker-code/NekoCore-OS "$T" && mkdir -p ~/.claude/skills && cp -r "$T/project/MA/MA-skills/search-archive" ~/.claude/skills/voardwalker-code-nekocore-os-search-archive && rm -rf "$T"
manifest:
project/MA/MA-skills/search-archive/SKILL.mdsource content
Search Archive
This skill enables searching the entity's long-term memory archive — memories older than 90 days that are no longer in the hot working-memory index.
When to Use
Use this skill when:
- The user asks about something from the distant past
- A relevant topic may be in older memories no longer appearing in recent context
- You want to cross-reference archived knowledge or past experiences
- You recall a memory exists but it's not surfacing in the current context
Do not use this for recent memories — those are retrieved automatically.
Tool Usage
[TOOL:search_archive query="your search terms here"]
With optional year range filter:
[TOOL:search_archive query="pipeline orchestration" yearRange='{"start":"2024-01-01T00:00:00.000Z","end":"2025-01-01T00:00:00.000Z"}' limit="10"]
Parameters
(required): Natural language search terms describing what you're looking forquery
(optional): JSON object withyearRange
and/or"start"
as ISO date strings to filter by archive date"end"
(optional): Number of results to return (default 5, max 20)limit
Result Format
Results include:
: Memory identifierid
: Extracted text content from the archived memorysummary
: When the memory was moved to the archivearchivedAt
: Topic keywords associated with the memorytopics
: Memory type (type
,episodic
, ordoc
)semantic_knowledge
: Relevance score (higher = more relevant)score
Behavior Guidelines
- Use concise, topic-focused queries — RAKE extraction handles phrase identification
- If results are returned, synthesize them naturally into your response
- Cite approximate time periods when referencing archived memories
- If no results are found, say so honestly — don't fabricate memories