Back to skill

Security audit

Canvas Claw

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent, but it handles passwords, tokens, prompts, and media with security gaps that should be reviewed before installation.

Install only if you trust the configured AI-video-agent endpoint and can use HTTPS or a strictly local localhost service. Avoid putting real passwords on the command line, protect any printed token, and do not submit sensitive prompts or media unless the remote service is approved for that data.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
canvas_claw/client.py:15
Finding

API credentials and user content can be transmitted over plaintext HTTP

Content
View full analysis
dict[str, str]: headers = { "token": self.config.token, "site-id": str(self.config.site_id), } if extra: headers.update(extra) return headers ``` ```python url = f"{self.config.base_url}{path}" if query: url = f"{url}?{parse.urlencode(query)}" req = request.Request( url, data=data, headers=self._build_headers(headers), method=method, ) ``` ```python def login(self, username: str, password: str) -> dict[str, Any]: payload = json.dumps( { "username": username, "password": password, } ).encode("utf-8") return self.request_json( "POST", "/api/login", data=payload, headers={"Content-Type": "application/json"}, ) ``` The base URL is loaded without validating or enforcing its scheme: ```python base_url = os.environ.get("AI_VIDEO_AGENT_BASE_URL", "").strip() ``` The installation instructions explicitly demonstrate a plaintext endpoint: ```bash export AI_VIDEO_AGENT_BASE_URL="http://localhost:8000" ``` ### Technical Analysis `CanvasClawClient` unconditionally attaches the API token and site identifier to requests. The login endpoint also sends the user's username and password in the HTTP request body. No check requires `AI_VIDEO_AGENT_BASE_URL` to use HTTPS. HTTP is reasonable for a strictly loopback-only development service, but the implementation does not limit HTTP to loopback addresses. A user can configure any remote HTTP host. When this occurs, credentials, API tokens, prompts, task information, and uploaded reference images or vid ...[truncated 1785 chars]
Remediation
View remediation
` header if supported by the service, while ensuring errors and logs never include the header value. 8. Add tests confirming that remote HTTP URLs are rejected, HTTPS URLs are accepted, and loopback HTTP is accepted only under the intended policy. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/login.py:17
Finding

Login credentials and session tokens are exposed through process arguments and standard output

Content
View full analysis
argparse.ArgumentParser: parser = argparse.ArgumentParser(description="Login to AI-video-agent") parser.add_argument("--base-url", required=True, help="AI-video-agent base URL") parser.add_argument("--site-id", required=True, type=int, help="Site ID") parser.add_argument("--username", required=True, help="Username") parser.add_argument("--password", required=True, help="Password") return parser def main() -> None: args = build_parser().parse_args() config = RuntimeConfig(base_url=args.base_url.rstrip("/"), token="", site_id=args.site_id) client = CanvasClawClient(config) session = login_and_extract_session( client, username=args.username, password=args.password, ) print(f"AI_VIDEO_AGENT_BASE_URL={args.base_url.rstrip('/')}") print(f"AI_VIDEO_AGENT_SITE_ID={args.site_id}") print(f"AI_VIDEO_AGENT_TOKEN={session.get('token', '')}") ``` The documented invocation also places the password directly on the command line: ```bash python3 ./scripts/login.py \ --base-url "http://localhost:8000" \ --site-id 10000 \ --username "demo" \ --password "secret" ``` ### Technical Analysis The login helper requires the password through an `argparse` command-line option. Depending on the operating system and environment, command-line arguments may be exposed through process inspection tools, process accounting, shell history, terminal logging, job-control systems, or automation logs. After authentication, the script prints the complete token to standard output. Standard output is frequently retained by terminal recorders, CI/CD systems, OpenClaw execution logs, shell redirection, or monitoring infrastructure. The token is t ...[truncated 1506 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (45)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill depends on environment-based runtime configuration and service connection parameters, but those data flows and operational semantics are not fully disclosed in the skill purpose. This matters because environment variables may contain sensitive tokens and service endpoints, and insufficient disclosure can lead to unsafe deployment or accidental credential leakage.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The markdown instructs users to pass a username and password directly on the command line, which can expose credentials through shell history or process listings. There is no accompanying warning about handling secrets safely or avoiding real credentials, so the guide omits an important user disclosure for a privacy-sensitive action.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The document instructs users to run npx clawhub login without pinning an exact package version. npx may fetch the latest package from the registry at execution time, which creates a supply-chain risk: a compromised or unexpected newer release could execute arbitrary code on the publisher's machine. In a publish workflow, this is more sensitive because the operator is likely authenticated and working in a trusted project directory.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The command npx clawhub whoami references an unpinned package, so running it can implicitly download and execute whatever version is currently published. That exposes users to supply-chain compromise and unexpected behavior, especially since this CLI is used around authentication state and developer identity verification.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The publish instruction uses npx clawhub publish without a pinned version, which means the release workflow depends on whatever package version npx resolves at runtime. If that package is malicious or compromised, it could steal tokens, alter the published artifact, or execute arbitrary code during a highly privileged deployment step, making this more dangerous than a read-only command.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
72% confidence
Finding

This duplicate finding points to the same unpinned npx clawhub login reference. The risk remains that readers may copy and execute a dynamically resolved package version, exposing their environment and credentials to a compromised CLI release.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The final instruction tells the user to continue with npx clawhub publish after login, again without pinning the package version. Because this is a deployment action performed after authentication, exploitation could lead to credential theft, malicious publication, or arbitrary code execution in a trusted developer context.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding

The skill declares capabilities that require environment access, file access, and network access, but it does not explicitly scope or document those permissions. This is dangerous because consumers and policy engines cannot easily determine the skill's effective trust boundary, increasing the chance of over-privileged execution and unintended data exposure to the external AI-video-agent service.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation does not warn users that prompts, reference images, and credentials are sent to an external AI-video-agent service. In this context, that omission is security-relevant because media prompts and uploaded files may contain sensitive personal or proprietary information, and users may not realize they are leaving the local trust boundary.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The function sends a username and password through client.login(...), which is a safety-relevant network/authentication operation involving credentials. This file provides no confirmation, logging, docstring, or comment disclosing that credential transmission occurs.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest describes a skill for generating and remixing images/videos through an AI video agent, but this client also exposes a dedicated /api/login capability that accepts raw username/password credentials. While network access to the generation service is expected, implementing an explicit login flow is additional behavior not reflected in the stated skill description.

Content

No source excerpt is available for this finding.