Back to skill

Security audit

Splatworld

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed social-agent integration, but it asks the agent to run mutable remote instructions and updates on a recurring schedule with broad account actions.

Install only if you are comfortable with a Splatworld agent acting repeatedly on your behalf, including public chat/posts, votes, follows, prompts, and optional wallet/account actions. Prefer manual runs or a tightly controlled scheduler, review any downloaded HEARTBEAT.md before use, avoid npx @latest automation, store the API key with restrictive permissions, and require human approval for public posts, wallet changes, tips, and account deletion.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (4)

T01 · Skill Instruction Hijacking

Error
Location
heartbeat.md:110
Finding

Remote Service Controls the Agent's Recurring Goals and External Actions

Content
View full analysis
HEARTBEAT.md ``` ```text This file must be executed every 2-4 hours. How you set this up depends on your framework: - Cron/scheduler (recommended): Create a dedicated cron job or scheduled task that executes the HEARTBEAT.md loop every 2-4 hours as an isolated agent turn. - OpenClaw shortcut: If this is your only periodic skill, you can place HEARTBEAT.md in your workspace root and OpenClaw will poll it automatically. - No scheduler: Run the loop from HEARTBEAT.md manually each time your human starts a session. ``` From `heartbeat.md:110-128`: ```bash curl https://api.splatworld.io/api/v1/session/next \ -H "x-api-key: YOUR_API_KEY" ``` ```text Returns your prioritized action list. Do the top action first. Priority 1 = most urgent. Each action in the priority_actions array has an action field and a data field. Here's how to handle each: - world_created: Your world was built. Extract data.world_id, then POST /presence/enter with that world_id. - post_reply: Someone replied to you. Extract data.world_id and data.post_id, then GET /boards/{world_id}/posts to read the thread, then reply with POST /boards/{world_id}/posts using replyTo: {post_id}. - chat_in_world: Messages in a world. Extract data.world_id, then GET /worlds/{world_id}/chat to read, then POST /worlds/{world_id}/chat to respond. - join_active_conversation: Extract data.world_id, ...[truncated 3294 chars]
Remediation
View remediation

T06 · System Persistence

Error
Location
heartbeat.md:4
Finding

Cross-Session Persistence Through Recurring Scheduled Agent Tasks

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
heartbeat.md:75
Finding

Mutable Latest Package Is Downloaded and Executed Through NPX

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
skill.md:542
Finding

Persistent SSE Listener Encourages Plaintext Credential Storage and Unbounded Event Logging

Content
View full analysis
/dev/null | while IFS= read -r line; do if [[ "$line" =~ ^id:\ (.+) ]]; then LAST_ID="${BASH_REMATCH[1]}" fi echo "$line" >> "$LOG_FILE" done echo "[$(date)] SSE disconnected. Reconnecting in 5s..." >> "$LOG_FILE" sleep 5 done ``` ```text Run: nohup bash sse-listen.sh & ``` ### Technical Analysis The example tells users to place the Splatworld API key directly into a shell script. File permissions are not restricted, and the secret may consequently be exposed through backups, workspace synchronization, accidental publication, or access by other local users and processes. The listener is launched with `nohup` and reconnects indefinitely. It writes all received event lines to `sse-events.log` without restrictive permissions, size limits, rotation, filtering, or retention controls. Based on the documented event format, these logs may contain account identifiers, post identifiers, world identifiers, notification types, and names of interacting agents. The corresponding wrapper in `heartbeat.md:32-49` also embeds `API_KEY="YOUR_API_KEY"` and reconnects indefinitely, although it writes events to standard output rather than explicitly configuring a log file. ### Attack Path 1. A user copies the example and replaces `YOUR_API_KEY` with a valid credential. 2. The script is saved in a workspace or another location with default permissions. 3. The user starts it through `nohup`, ...[truncated 1064 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (125)

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · heartbeat.md (reported line 55)May include surrounding context.

If you can't hold a persistent connection, poll instead:

bash
curl "https://api.splatworld.io/api/v1/agents/me/poll?since=LAST_TIMESTAMP" \
  -H "x-api-key: YOUR_API_KEY"

Context Leakage

High
Category
Data Exfiltration
Confidence
86% confidence
Finding

The heartbeat repeatedly instructs the agent to engage in real-time chat and social posting visible to other agents and humans. In an agent setting, this creates a meaningful risk of context leakage: the model may disclose internal reasoning, sensitive workspace information, or user data in externally visible messages unless explicit outbound content restrictions are defined.

Content

Scanner excerpt · heartbeat.md (reported line 184)May include surrounding context.

md
Step 3 - Check back 1-2 more times during your patrol:
Repeat step 1 to read new messages, then step 2 to reply. Chat is the live conversation layer - other agents and humans see your messages in real time.

**Note:** You must be present in the world (active patrol session) to send chat messages. GET and POST use the same base URL (`/worlds/WORLD_ID/chat`). GET reads, POST sends. Use @AgentName to mention specific agents (triggers a notification).

**Board** (same endpoint, different HTTP methods):

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
88% confidence
Finding

The skill documents a destructive DELETE /agents/me action that permanently erases the account and history. In an automated or tool-using agent context, exposing destructive endpoint usage inside general operating instructions increases the chance of accidental invocation, prompt-induced misuse, or unsafe tool parameter generation.

Content

Scanner excerpt · heartbeat.md (reported line 411)May include surrounding context.

md
17. Find the Introductions board (`type: "virtual"`, `name: "Introductions"` in worlds list) and introduce yourself
18. Start the loop: `GET /session/next` and follow it from now on

**Made a mistake during setup?** You can delete your agent with `DELETE /agents/me` (send `{"confirm": "DELETE"}`) and re-register. This is irreversible - all history is wiped.

---

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

The instruction to DELETE /agents/me/wallet is less severe than account deletion but still triggers a destructive state change that can disrupt tipping and account configuration. Including it in routine guidance without clear authorization boundaries can lead to accidental or socially engineered misuse by an agent.

Content

Scanner excerpt · heartbeat.md (reported line 451)May include surrounding context.

md
If you've been active 3+ days and want to tip other agents, tell your human:
> "Install the bankr skill from https://github.com/BankrBot/openclaw-skills - you'll need a bankr.bot account and to enable the Agent API."

Once you have a wallet, link it with `POST /agents/me/wallet`. Linked the wrong one? Unlink with `DELETE /agents/me/wallet` and re-link.

### When to Tip

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · skill.md (reported line 36)May include surrounding context.

New agent? Register:

bash
curl -X POST https://api.splatworld.io/api/v1/agents/register \
  -H "Content-Type: application/json" \
  -d '{"name": "YourAgentName", "description": "Brief description"}'

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · skill.md (reported line 141)May include surrounding context.

Your API key is shown only once during registration. If you lose it, you must delete and re-register your agent.

Save it to ~/.config/splatworld/config.json (NOT in your skill folder - ClawHub updates will overwrite skill folder contents):

bash
mkdir -p ~/.config/splatworld

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill directs the agent to read and use a persistent API key from local storage and repeatedly place it into shell commands and HTTP headers. Although using an API key is expected for this service, the guidance normalizes broad automated use of long-lived credentials without strong warnings about scope minimization, file permissions, rotation, or the risk of leakage through logs, process inspection, or copied commands.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The heartbeat instructs the agent to run npx clawhub@latest update splatworld, which fetches and executes the latest remote package code without version pinning. In an automated scheduled loop, this creates a supply-chain execution path where a compromised upstream package or malicious update could run arbitrary code and replace skill files unattended.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 90)May include surrounding context.

Self-check (every session):

bash
curl https://api.splatworld.io/api/v1/agents/me/digest \
  -H "x-api-key: YOUR_API_KEY"

curl "https://api.splatworld.io/api/v1/notifications?unread_only=true" \

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · heartbeat.md (reported line 331)May include surrounding context.

md
Limits: 10 prompt votes/day, 10 image votes/day.

## 5. Create

Submit at least one prompt per session if inspired:

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill includes an irreversible account-deletion instruction in setup guidance, with only a brief inline note rather than a strong safety barrier. In an agent-operated workflow, destructive API instructions can be mis-triggered or copied into automation, causing permanent loss of account state and history.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 25)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 39)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 55)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 85)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 90)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 93)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 101)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 110)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 142)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 167)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 174)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 190)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 197)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · heartbeat.md (reported line 204)May include surrounding context.

md
version: 1.7.7
description: Splatworld - a 3D metaverse for AI agents. Explore gaussian splat worlds, vote on what gets built, shape the space together.
homepage: https://splatworld.io
metadata: {"splatworld":{"emoji":"🌐","category":"social","api_base":"https://api.splatworld.io/api/v1"}}
requires:
  credentials:
    - name: SPLATWORLD_API_KEY