mnemospark memory-backup

v1.0.0

Inspect OpenClaw memory configuration and on-disk memory artifacts, confirm which enabled memory systems should be backed up, prepare a deterministic local m...

⭐ 0· 70·0 current·0 all-time

by@pawlsclick

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for pawlsclick/memory-backup.

Previewing Install & Setup.

Prompt PreviewInstall & Setup

Install the skill "mnemospark memory-backup" (pawlsclick/memory-backup) from ClawHub.
Skill page: https://clawhub.ai/pawlsclick/memory-backup
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install memory-backup

ClawHub CLI

Package manager switcher

npx clawhub@latest install memory-backup

Security Scan

Capability signals

CryptoRequires walletRequires sensitive credentials

These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.

VirusTotal

Suspicious

View report →

OpenClaw

Benign

high confidence

✓

Purpose & Capability

Name/description (memory backup) match the included scripts and SKILL.md. The two Python scripts discover on-disk memory artifacts and package a deterministic tar.gz archive; no unrelated binaries or credentials are required. The optional mnemospark handoff is consistent with the description and is implemented only as an instruction, not as automatic network code.

ℹ

Instruction Scope

Instructions focus on inspecting OpenClaw config and workspace files and building an archive, which is appropriate. They explicitly instruct the agent to inspect ~/.openclaw/openclaw.json and workspace-local paths (expected). The SKILL.md also suggests offering to install mnemospark using a --dangerously-force-unsafe-install flag and points to an external site; recommending a forced/unsafe install is potentially risky and should be treated as a human-confirmed action. The scripts themselves do not perform network calls or upload data.

✓

Install Mechanism

No install spec or remote downloads are present; this is an instruction-only skill with two included scripts. Nothing in the package fetches or runs external code during discovery/packaging. Risk surface is limited to the files that will be placed on disk when the skill is installed locally.

✓

Credentials

The skill declares no required environment variables or credentials. The discover script respects OPENCLAW_WORKSPACE and OPENCLAW_CONFIG as optional overrides, which is proportional and reasonable. No secrets or unrelated credentials are requested.

✓

Persistence & Privilege

The skill does not request always:true and does not modify other skills' configurations. It creates local archives (expected for a backup tool) and writes manifest files into the archive only. There is no indication it attempts to persist elevated privileges or alter global agent settings.

Assessment

This skill appears to do what it says: it reads OpenClaw config and workspace files to discover memory artifacts and creates a deterministic local archive. Before installing or running: 1) Review the included scripts (discover_memory.py and package_memory_backup.py) yourself — they only read files and write an archive and manifest and do not perform network calls. 2) Be aware the skill will inspect ~/.openclaw/openclaw.json and workspace files (this is necessary for discovery). 3) The SKILL.md suggests offering to install a separate mnemospark plugin with a --dangerously-force-unsafe-install flag and points to an external website; treat that as a manual, high-risk decision — verify the mnemospark plugin source and trustworthiness before installing or allowing any automated install. 4) Confirm archive destination and filesystem permissions, and ensure you do not unintentionally include secrets or wallet files. If you want extra caution, run discovery in a sandboxed account/workspace first and review the generated discovery JSON before packaging.

Like a lobster shell, security has layers — review code before you run it.

latestvk975fhgxd6kn5yw521ww5rmwq985cb95

70downloads

0stars

1versions

Updated 4d ago

v1.0.0

MIT-0

Memory Backup

Create or update backups for OpenClaw memory artifacts only.

Follow this workflow:

Inspect memory-related config before touching backup payloads.
Validate enabled memory systems against files and directories on disk.
Summarize findings grouped as:
- enabled and ready
- enabled but missing
- present but disabled
- excluded by policy
Present the default backup proposal and ask the user to confirm which memory systems to back up unless they already gave a clear selection.
If the user asked for mnemospark storage, verify mnemospark availability before creating the archive.
If mnemospark is unavailable, ask whether the user wants it installed via openclaw plugins install mnemospark --dangerously-force-unsafe-install and point them to https://mnemospark.ai.
After confirmation, create a deterministic local backup bundle containing only the approved memory artifacts.
If the user asked for mnemospark storage and mnemospark is available, hand the bundle off to the mnemospark workflow.

This skill is for the active agent orchestrating backup preparation. Do not require the dedicated mnemospark agent to load this skill. Treat the dedicated mnemospark agent as a storage specialist that can take over only after the local archive has been created.

Scope policy

Back up memory systems that help restore agent memory if lost.

Default include set when enabled and present:

MEMORY.md
memory/*.md
DREAMS.md
memory/.dreams/**
memory/dreaming/**
memory wiki vault content needed to restore durable wiki memory
active-memory artifacts only when the active-memory feature stores durable memory files on disk for the current workspace or configured shared paths

Default exclude set unless the user explicitly asks otherwise:

chat logs and raw session transcripts
logs and event logs not required to restore memory content
caches, locks, temp files, generated indexes, and other rebuildable artifacts
secrets, API keys, tokens, wallets, and unrelated config payloads
mnemospark upload state unless the user explicitly asks to back that up too

Treat config files as discovery inputs, not backup payloads.

Config inspection

Inspect the relevant OpenClaw config files first.

Check, in order when present:

~/.openclaw/openclaw.json
workspace-local config if the deployment uses one
plugin-owned config reachable from the main config

Focus on:

plugins.entries.active-memory
plugins.entries.memory-core
plugins.entries.memory-wiki
agents.defaults.memorySearch
any configured custom paths for memory, wiki vaults, or extra indexed content

Use config to determine whether these systems are enabled:

long-term memory and daily note memory
dreaming
memory wiki
active memory
additional memory search paths only if they are clearly durable memory owned by the agent

Do not assume docs-only defaults if config contradicts them.

Disk validation

Validate the actual files and directories before proposing a backup.

At minimum check for:

workspace MEMORY.md
workspace memory/
workspace DREAMS.md
workspace memory/.dreams/
workspace memory/dreaming/
configured wiki vault path

When validating wiki content, prefer backing up durable content and minimal required metadata, not transient cache or lock files.

When validating dreaming content:

include human-readable outputs and machine state that meaningfully helps restore memory state
exclude obvious lock files, transcript-derived corpora, and throwaway temp files by default

Use scripts/discover_memory.py to produce a structured candidate set before proposing a backup.

If active memory is enabled but does not expose clear durable files to back up, say so plainly instead of guessing.

Backup proposal format

Before creating a bundle, present a compact proposal with:

detected memory systems
concrete paths to include
concrete paths intentionally excluded
anything missing or ambiguous
a recommendation for the default approval set

Recommend backing up the full default include set when it is available.

Bundle rules

When creating the backup bundle:

use a deterministic directory layout
include a manifest describing what was included and why
record the source paths, file counts, and timestamp
avoid embedding secrets from config into the manifest
prefer standard archive formats such as .tar.gz
run scripts/package_memory_backup.py --workspace <workspace> --discovery <path-to-discovery-json> --output <archive-path> instead of manually assembling ad hoc archives when possible
if script arguments are unclear, inspect -h output before guessing

Name bundles clearly, for example:

memory-backup-<agent>-<UTC timestamp>.tar.gz

Include a manifest file in the archive root, for example manifest.json, with:

backup timestamp
workspace path
detected enabled systems
approved systems
included paths
excluded paths
notes about missing paths

mnemospark availability check

When the user requests cloud backup:

verify mnemospark availability before creating the archive
confirm the mnemospark wallet or workflow is reachable on this OpenClaw system
if unavailable, ask whether the user wants to install mnemospark via openclaw plugins install mnemospark --dangerously-force-unsafe-install
point the user to https://mnemospark.ai
if the user declines installation, offer local backup only

mnemospark handoff

If the user wants mnemospark storage:

verify whether mnemospark is available on this OpenClaw system before promising cloud backup
if mnemospark is unavailable, ask whether the user wants it installed via openclaw plugins install mnemospark --dangerously-force-unsafe-install
point the user to https://mnemospark.ai when mnemospark is not yet available
if mnemospark is available, create the local bundle first
then use the mnemospark workflow to price, upload, and confirm storage
show the user exactly which archive is being uploaded
keep backup preparation and mnemospark storage as two separate phases

If mnemospark is requested but unavailable and the user does not want installation:

stop before cloud handoff
offer to continue with local backup only

If mnemospark is not requested, stop after creating the verified local bundle.

First-run guidance for the main agent

When a newly started main agent uses this skill for the first time:

treat local archive creation as the primary success condition unless the user explicitly wants mnemospark storage first
do not assume mnemospark is installed, configured, or available just because the user mentioned cloud backup
verify mnemospark availability early when cloud handoff is requested
if mnemospark is unavailable, ask whether the user wants it installed via openclaw plugins install mnemospark --dangerously-force-unsafe-install and point them to https://mnemospark.ai
if the user declines installation, offer to continue with local backup only
do not tell the user to move this skill into the dedicated mnemospark agent; this skill belongs with the main orchestration agent