Back to skill

Security audit

yunitalk-beta

Security checks for vulnerabilities and agentic risk

Overview

The skill is a real Talk Robots/OpenClaw integration, but it asks users to run unpinned npm code that reads local OpenClaw tokens and its send script prints signed request details.

Install only if you trust the publisher and the npm package source. Prefer a pinned, reviewed version of `yntalk-openclaw-init`, keep the generated QR/init string private because it contains the OpenClaw token, restrict `TALK_ROBOT_ENDPOINT` to a trusted Talk Robots endpoint, and avoid capturing shell output in shared logs because it includes signed request details.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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 (22)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared description frames the skill as a chat integration, but the documented behavior includes authenticated HTTP POSTs, file uploads, and message dispatch to arbitrary or configurable endpoints using sensitive credentials. This mismatch can mislead reviewers and users about the true operational scope, reducing scrutiny around capabilities that could be abused for data exfiltration, unauthorized messaging, or SSRF-like access to internal endpoints via configurable URLs.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 68)May include surrounding context.

md
**方式二:`shell/robot_send.sh`**。clawhub 发布包因大小限制不包含 `bin/`;如果没有 `bin/bot-cli`,使用内置 shell 脚本(依赖 `curl`、`openssl` 和 `python3`,同样自动完成签名):

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 71)May include surrounding context.

md
**方式二:`shell/robot_send.sh`**。clawhub 发布包因大小限制不包含 `bin/`;如果没有 `bin/bot-cli`,使用内置 shell 脚本(依赖 `curl`、`openssl` 和 `python3`,同样自动完成签名):

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 77)May include surrounding context.

md
**方式二:`shell/robot_send.sh`**。clawhub 发布包因大小限制不包含 `bin/`;如果没有 `bin/bot-cli`,使用内置 shell 脚本(依赖 `curl`、`openssl` 和 `python3`,同样自动完成签名):

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 87)May include surrounding context.

md
**方式二:`shell/robot_send.sh`**。clawhub 发布包因大小限制不包含 `bin/`;如果没有 `bin/bot-cli`,使用内置 shell 脚本(依赖 `curl`、`openssl` 和 `python3`,同样自动完成签名):

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

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

md
**方式二:`shell/robot_send.sh`**。clawhub 发布包因大小限制不包含 `bin/`;如果没有 `bin/bot-cli`,使用内置 shell 脚本(依赖 `curl`、`openssl` 和 `python3`,同样自动完成签名):

Rp1

Medium
Category
MCP Rug Pull
Confidence
87% confidence
Finding

The README instructs users to run npx yntalk-openclaw-init without pinning a package version. That causes execution of whatever package version is current at install time, creating a supply-chain risk if the npm package is later compromised, typosquatted, or updated maliciously. In this skill's context, the CLI reads local openclaw.json, extracts gateway tokens, and generates initialization data, so a compromised package could directly exfiltrate sensitive OpenClaw credentials.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
84% confidence
Finding

This duplicate finding points to the same unpinned npx yntalk-openclaw-init usage in the troubleshooting guidance. Executing an unpinned npm package can run attacker-controlled code if the package or dependency chain is compromised, which is especially relevant here because the CLI handles local service endpoints and authentication tokens.

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 · ROBOT_SEND_API.md (reported line 187)May include surrounding context.

curl 示例:

bash
curl -X POST "$request_url" \
  -H "Content-Type: application/json" \
  --data-binary '{
    "cmd": "msg",

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 · ROBOT_SEND_CURL_API.md (reported line 49)May include surrounding context.

msg_id=<msg_id>&noncestr=&robot_id=<robot_id>&timestamp=

text

## curl 发送文本消息

文本消息使用 `application/json`,请求体是完整 `imx.Msg` JSON 结构,`cmd` 固定为 `msg`,文本内容放在 `body.content`。

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill explicitly instructs use of shell commands and outbound network operations, but it declares no corresponding tool scope or permission boundaries. That makes the skill harder to review and easier to over-grant at runtime, increasing the chance that an agent can invoke shell/network capabilities without clear user or platform visibility.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
87% confidence
Finding

Using npx yntalk-openclaw-init without a pinned version pulls code dynamically from the package registry at execution time, creating a supply-chain risk. If the package is compromised, unpublished/replaced, or unexpectedly updated, arbitrary code could run on the host and expose secrets such as OpenClaw connection data or robot credentials.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Line L138 sets the default language to "zh-CN", which imposes a specific locale in the manifest. The file does not indicate any user opt-in, alternative language selection, or region-specific justification for this constraint.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The README instructs users to run npx yntalk-openclaw-init without pinning a specific package version. This makes execution dependent on whatever version is currently published to the registry at runtime, creating a supply-chain risk if a malicious or compromised release is later published.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

This usage example again tells users to execute the package via npx without a pinned version. An attacker who gains control of the package or publishes a malicious update could cause users following the documentation to run unintended code.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The command example uses npx with no version constraint, so it will resolve to the latest registry version at time of execution. In this skill's context, that is more sensitive because the tool reads local openclaw.json and handles gateway tokens, so a malicious package update could exfiltrate credentials or network details.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The help example npx yntalk-openclaw-init -h still requires fetching and executing the package, and without version pinning it inherits the same supply-chain risk. Even though -h appears harmless, installation-time or startup code would still run before help output is shown.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

This example combines an environment variable containing a local config path with an unpinned npx execution. Because the tool is documented as reading local OpenClaw configuration and tokens, an untrusted future package version could access sensitive files and secrets from the specified path.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script unconditionally prints the full request URL, signing text, and computed signature before sending. While it does not print the raw robot_key, these values are credential-derived and can leak sensitive operational data into terminal scrollback, CI logs, shell history captures, or centralized logging systems, increasing the risk of replay attempts or unauthorized message submission if the server's verification controls are weak.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 · shell/robot_send.sh (reported line 244)May include surrounding context.

sh
if [[ "${mode}" == "text" ]]; then
  body="$(json_body)"
  curl -sS -X POST "${request_url}" \
    -H "Content-Type: application/json" \
    --data-binary "${body}" \
    -w '\nHTTP_STATUS=%{http_code}\n'

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill documentation is entirely written in Chinese from the title onward, and the file does not indicate that Chinese is optional, user-selectable, or required for a specific regional/compliance reason. Under the stated policy, forcing a single language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.