Back to skill

Security audit

Text to Infographic · 文字转信息图

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report unrelated online registry checks, weak numerical validation, and an automatic browser sandbox downgrade despite its stated infographic purpose.

ClawScan places this skill in Review, with medium overall confidence, and its scanner recommends not installing. Before considering it, review the screenshot sandbox fallback, the online verification behavior, and the numerical source check; do not treat the completed scans as proof that the skill is safe.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
Findings (12)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

该代码的主要用途是发布维护与版本同步审计,而非描述中的信息图构建、校验或渲染。它校验的是软件版本和脚本指纹,不是信息图 plan 的布局语义或数值一致性;此外还默认查询线上技能仓库。这些属于描述未披露的独立维护能力,而不是所声明信息图处理流程的实现细节。未发现代码自行发布、安装、写文件或上传本地文件内容的行为。

Content

No source excerpt is available for this finding.

Code capability 'env' detected in scripts/export_png.py but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

L074–L078 invoke a network-backed ClawHub verification command unrelated to the stated infographic-generation purpose.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest describes text-to-plan conversion, validation, local HTML rendering, and optional screenshots using an existing browser; it does not declare remote package-registry checks. This script invokes clawhub skill verify for a hard-coded registry package, and L099–L102 enable that lookup by default unless --offline is supplied. This is a release-maintenance capability rather than a requirement of infographic generation.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/check_sync.py (reported line 77)May include surrounding context.

python
def read_online():
    """线上 ClawHub 状态(依赖 clawhub CLI + 网络)。"""
    try:
        r = subprocess.run(["clawhub", "skill", "verify", SKILL],
                           capture_output=True, text=True, timeout=30)
        d = json.loads(r.stdout)
        created = d.get("createdAt")

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.

Content

Scanner excerpt · scripts/export_png.py (reported line 171)May include surrounding context.

python
# ---------------------------------------------------------------- 截图
def _shoot(chrome, html_path, out_png, width, height, scale, no_sandbox=False):
    url = "file://" + os.path.abspath(html_path)
    cmd = [chrome, "--headless=new", "--disable-gpu", "--hide-scrollbars",
           "--disable-extensions", "--no-first-run", "--no-default-browser-check",

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/export_png.py (reported line 182)May include surrounding context.

python
cmd.insert(2, "--no-sandbox")
    if os.path.exists(out_png):
        os.remove(out_png)
    proc = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, timeout=90)
    if not os.path.exists(out_png) or os.path.getsize(out_png) == 0:
        raise ScreenshotFailed(
            f"截图失败(退出码 {proc.returncode}):\n"

Tainted flow: 'cmd' from os.environ.get (line 173, credential/environment) → subprocess.run (code execution)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/export_png.py (reported line 182)May include surrounding context.

python
cmd.insert(2, "--no-sandbox")
    if os.path.exists(out_png):
        os.remove(out_png)
    proc = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, timeout=90)
    if not os.path.exists(out_png) or os.path.getsize(out_png) == 0:
        raise ScreenshotFailed(
            f"截图失败(退出码 {proc.returncode}):\n"

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.

Content

Scanner excerpt · scripts/export_png.py (reported line 194)May include surrounding context.

python
chrome = find_chrome(chrome)
    tmpdir = tempfile.mkdtemp(prefix="ig-png-")
    probe = os.path.join(tmpdir, "probe.png")
    no_sandbox = False

    def shoot(target, h, s, note=""):
        """截图,失败时降级 --no-sandbox 重试一次(容器/CI 里 Chrome 常起不来)。

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
98% confidence
Finding

The surrounding exception handler automatically retries any screenshot failure with --no-sandbox and retains that setting for subsequent screenshots. This silently removes browser process isolation when rendering potentially untrusted HTML; exploiting a browser vulnerability could then compromise the export process's host privileges, although renderer-generated static HTML reduces exposure.

Content

Scanner excerpt · scripts/export_png.py (reported line 201)May include surrounding context.

python
一旦降级成功,本次运行后续截图都直接带 --no-sandbox,不再重复试错。"""
        nonlocal no_sandbox
        try:
            _shoot(chrome, html_path, target, width, h, s, no_sandbox=no_sandbox)
        except ScreenshotFailed:
            if no_sandbox:
                raise

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/test_export_png.py (reported line 98)May include surrounding context.

python
html = os.path.join(here, "examples", "dashboard-demo.json")
    import subprocess
    out_html = os.path.join(tmp, "dash.html")
    subprocess.run([sys.executable, os.path.join(here, "scripts",
                   "render_infographic_html.py"), html, out_html],
                   check=True, stdout=subprocess.DEVNULL)
    out_png = os.path.join(tmp, "dash.png")

Numerical source validation accepts unsupported values merely because one digit appears in the source.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The provenance check suppresses a warning whenever any single digit of a proposed value occurs anywhere in the source, so a fabricated value such as 9000 can pass against unrelated text containing 9. This bypasses the advertised numerical-consistency control, including strict mode because no warning is generated, and can allow unsupported figures into trusted infographics.

Content

No source excerpt is available for this finding.

The --explain feature specifies Chinese-only diagnostics without offering a language choice.

Low
Category
Output Handling
Confidence
92% confidence
Finding

The specification explicitly requires Chinese diagnostic text for the why/fix fields, including JSON output. It provides neither a user-selectable language nor a region-specific justification; opting into explanations does not explicitly constitute opting into Chinese.

Content

No source excerpt is available for this finding.