How do I install the Curate skill?

Run npx skillmds@latest add study8677/curate in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Curate safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Curate?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Curate free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Curate?

study8677 (@study8677) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Curate by study8677 — Agent Skill

Name: Curate
Author: study8677

File contents

Curate MemoBox Memory

Use curation commands to keep an opted-in .memobox useful and lightweight. Review .memobox/index.json directly before modifying records. MemoBox may expose duplicate candidates, but the model or user makes every merge, stale, pin, archive, and promotion decision.

Do not initialize a missing store. Do not curate during simple or sensitive tasks unless the user explicitly asks. A weekly review during dogfooding is usually enough.

Review Order

Read .memobox/index.json first.
Review needs_review records created by the PreCompact(auto) safety net; keep their durable content in a deliberate Memory Mail or archive/stale-mark the checkpoint.
Open candidate bodies before merging, marking stale, pinning, or promoting.
Prefer fewer high-value records over transcript-like accumulation.
Preserve source_refs such as memobox:<id> because they make real reuse measurable.

Common Commands

Find likely duplicates:

memobox --store .memobox curate duplicates --json

Merge duplicates and archive the source memories:

memobox --store .memobox curate merge <id-a> <id-b> \
  --subject "<merged subject>" \
  --summary "<merged index summary>" \
  --json

Mark exact memory ids as stale after reviewing the index:

memobox --store .memobox status <id-a> stale
memobox --store .memobox status <id-b> stale

Pin exact memory ids after reviewing the index:

memobox --store .memobox status <id-a> pinned
memobox --store .memobox status <id-b> pinned

Promote reusable project memory into global memory only after confirming that the body is verified, portable beyond the current workspace, understandable without private repository context, and free of sensitive data:

memobox --store .memobox promote <memory-id> \
  --global-store "${MEMOBOX_GLOBAL_STORE:-$HOME/.memobox-global}" \
  --tag "<reusable-pattern>" \
  --json

Before merging, stale-marking, pinning, or promoting, read .memobox/index.json and summarize the exact ids you selected unless the user already gave exact ids and intent. The duplicates command is only a candidate list; it is not a relevance or merge decision.

Keep the global index intentionally small. Do not promote one-off project status, repository-specific paths, speculative conclusions, secrets, personal data, or confidential raw traces. When global knowledge is reused in a project, preserve it with --source-ref "memobox-global:<id>".

---name: curatedescription: Use when the user asks to curate, deduplicate, merge, pin, archive, mark stale, promote, or clean up MemoBox memories.---# Curate MemoBox MemoryUse curation commands to keep an opted-in `.memobox` useful and lightweight. Review `.memobox/index.json` directly before modifying records. MemoBox may expose duplicate candidates, but the model or user makes every merge, stale, pin, archive, and promotion decision.Do not initialize a missing store. Do not curate during simple or sensitive tasks unless the user explicitly asks. A weekly review during dogfooding is usually enough.## Review Order1. Read `.memobox/index.json` first.2. Review `needs_review` records created by the `PreCompact(auto)` safety net; keep their durable content in a deliberate Memory Mail or archive/stale-mark the checkpoint.3. Open candidate bodies before merging, marking stale, pinning, or promoting.4. Prefer fewer high-value records over transcript-like accumulation.5. Preserve `source_refs` such as `memobox:<id>` because they make real reuse measurable.## Common CommandsFind likely duplicates:```bashmemobox --store .memobox curate duplicates --json```Merge duplicates and archive the source memories:```bashmemobox --store .memobox curate merge <id-a> <id-b> \ --subject "<merged subject>" \ --summary "<merged index summary>" \ --json```Mark exact memory ids as stale after reviewing the index:```bashmemobox --store .memobox status <id-a> stalememobox --store .memobox status <id-b> stale```Pin exact memory ids after reviewing the index:```bashmemobox --store .memobox status <id-a> pinnedmemobox --store .memobox status <id-b> pinned```Promote reusable project memory into global memory only after confirming that the body is verified, portable beyond the current workspace, understandable without private repository context, and free of sensitive data:```bashmemobox --store .memobox promote <memory-id> \ --global-store "${MEMOBOX_GLOBAL_STORE:-$HOME/.memobox-global}" \ --tag "<reusable-pattern>" \ --json```Before merging, stale-marking, pinning, or promoting, read `.memobox/index.json` and summarize the exact ids you selected unless the user already gave exact ids and intent. The duplicates command is only a candidate list; it is not a relevance or merge decision.Keep the global index intentionally small. Do not promote one-off project status, repository-specific paths, speculative conclusions, secrets, personal data, or confidential raw traces. When global knowledge is reused in a project, preserve it with `--source-ref "memobox-global:<id>"`.

Frequently asked questions

Curate