Openclaw Spacesuit

Security checks across malware telemetry and agentic risk

Overview

This workspace framework is mostly coherent, but it gives the agent broad authority to search private files and credentials, persist and push memory, and open tunnels without clear user approval.

Install only if you want a highly opinionated workspace framework. Before using it, narrow the search and credential rules, decide whether memory files may be committed or pushed, and require explicit approval before reading secrets, scanning private folders, or starting public tunnels.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Risk analysis

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

#
ASI02: Tool Misuse and Exploitation
Medium
What this means

The agent may inspect private cloud-synced files, dotfiles, prior sessions, or memory without a clear per-task approval boundary.

Why it was flagged

The skill instructs the agent to search broad local and synced private areas, not just the workspace, when trying to find information.

Skill content
EXHAUST ALL OPTIONS before saying "not found": ... Cloud storage / synced drives ... Home config: `~/.config/`, dotfiles ... Sessions: Recent history, handoffs ... Memory: Semantic search
Recommendation

Limit search instructions to user-approved workspace paths by default and require explicit confirmation before searching home config, cloud storage, prior sessions, or other private locations.

#
ASI03: Identity and Privilege Abuse
High
What this means

The agent could access API keys, tokens, or other secrets beyond what a user expected this scaffold skill to use.

Why it was flagged

The skill directs the agent to search common credential stores broadly, while the registry declares no credential requirement or scoped credential contract.

Skill content
When looking for API keys, tokens, or credentials: 1. Workspace `.envrc` ... 2. Workspace `.env` ... 3. Gateway config ... 4. Tool-specific config — `~/.config/{tool}/` ... 5. Environment variables — `env | grep -i {keyword}` ... ALWAYS check .envrc first.
Recommendation

Declare any expected credential use, restrict lookup to named services needed for a task, and require user approval before reading secret-bearing files or environment variables.

#
ASI06: Memory and Context Poisoning
Medium
What this means

Personal context may be stored long-term, reused by future sessions, and pushed outside the machine if the workspace has a git remote.

Why it was flagged

The framework creates persistent personal memory and encourages committing and pushing those notes, which can retain sensitive context and sync it to a remote repository.

Skill content
You can **read, edit, and update** MEMORY.md freely in main sessions ... Write significant events, thoughts, decisions, opinions, lessons learned ... `git add memory/*.md && git commit -m "memory: $(date +%Y-%m-%d) notes" && git push`
Recommendation

Add explicit retention rules, review-before-write and review-before-push steps, and clear exclusions for sensitive personal data.

#
ASI02: Tool Misuse and Exploitation
Medium
What this means

A local development service could be exposed through a tunnel and shared more broadly than intended.

Why it was flagged

The instruction encourages automatically opening and maintaining an externally reachable tunnel during web UI development without requiring explicit user approval.

Skill content
When developing web UIs ... **Always start with tunnel** — `./start.sh --tunnel` or equivalent ... **Always share the link** ... **Keep it running**
Recommendation

Require explicit user confirmation before starting tunnels, document what will be exposed, and stop tunnels by default when work is complete.

#
ASI06: Memory and Context Poisoning
Low
What this means

Running the script can create a local index of people from prior Slack/session transcripts.

Why it was flagged

The utility script is purpose-aligned and user-invoked, but it reads local session transcripts and stores Slack operator identifiers and message counts.

Skill content
Scans OpenClaw session transcripts and extracts unique users who have interacted with the bot. Outputs to state/operators.json.
Recommendation

Use `--dry-run` first, review `state/operators.json`, and avoid sharing that file if operator identities are private.

#
ASI05: Unexpected Code Execution
Low
What this means

Running the installer changes the workspace root and adds executable scripts.

Why it was flagged

The setup script writes workspace files, creates directories, and copies executable utility scripts; this is expected for the scaffold but is still local code execution.

Skill content
mkdir -p "$WORKSPACE/memory" ... mkdir -p "$WORKSPACE/scripts" ... cp "$script" "$target" ... chmod +x "$target" ... echo "$VERSION" > "$WORKSPACE/.spacesuit-version"
Recommendation

Review the generated files after install, run setup from the intended workspace only, and prefer dry-run or diff commands before upgrades.