Back to skill

Security audit

schematic-reader

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because scanners report purpose and behavior mismatches, heuristic review judgments, and installation and execution risks; no clear malicious intent is established.

Consider avoiding installation unless you need these capabilities and have reviewed the reported concerns. Before using it, verify whether ERC triage could hide findings, whether PDF probing or optional package installation is needed, and whether you trust the unpinned external MCP package and package mirror. Treat scan coverage as incomplete: Skillspector reports partial analysis because some local references were unresolved, so the results do not establish that unreported behavior is absent.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/lceda_parse.py
Finding

Untrusted project metadata can disclose schematic data outside the selected project

Content
View full analysis

Impact

An attacker-supplied project can cause confidential schematic information from another accessible project to enter the Agent context or generated report. This is format-constrained local disclosure; arbitrary file-content disclosure and network exfiltration are not established.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (23)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The mismatch is in triage_single_pin() and the erc --triage command. Nets or pins containing 'unconnected' are classified as ignored; single-pin nets attached to selected reference prefixes are characterized as normal module/connector cases; remaining findings are labeled as needing review. These judgments rely on names and prefixes rather than establishing electrical correctness. Triage output also replaces the full ERC result with these single-pin buckets. Although deterministic, this classification goes beyond factual ERC enumeration and contradicts the declared no-review boundary. External parser execution, optional npx package retrieval and netlist export are supporting implementation details, not independent mismatches.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied script's primary output is a PDF text-layer classification and token inventory, not a parsed electrical netlist. PDF is not among the declared input formats, and the script explicitly states that connectivity cannot be recovered. Although PDF probing could support a broader schematic-processing workflow, this chunk implements a materially different input/output capability from the description. The optional installation path also accesses an external package mirror and modifies the user Python environment; this occurs only when the dependency is missing and --install-pymupdf is supplied. There is no evidence of exfiltration or malicious intent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The supplied commands operate on a PDF rather than any declared schematic or netlist format. PDF inspection is related to schematic analysis, but the visible chunk exposes an undeclared input capability and does not demonstrate the promised pin-level netlist parsing or connectivity/ERC operations. The script implementation is not supplied, so its exact extraction behavior cannot be verified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
82% confidence
Finding

Most listed commands are consistent with schematic parsing and netlist inspection. The explicit addition of single-pin ERC triage, however, indicates interpretation or classification of findings rather than merely enumerating deterministic parsing facts. This conflicts with the description's stated boundary against review judgments. Only command examples are supplied, so the implementation and exact nature of that triage cannot be verified; no unrelated resource access or malicious behavior is demonstrated.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
82% confidence
Finding

The visible commands focus on design variants rather than the declared schematic parsing, connectivity queries, or ERC enumeration. Variant comparison is an additional user-facing capability, not merely a demonstrated parsing implementation detail. The implementation of un_mcp.py is absent, so the precise meaning of vdns and any underlying file access or side effects cannot be established.

Content

No source excerpt is available for this finding.

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · scripts/run_regression.py (reported line 47)May include surrounding context.

python
def run(name, cmd, expect_exit=0, must_contain=None, must_not_contain=None,
        env=None):
    """跑一条命令并断言 exit code 与输出内容。"""
    full_env = dict(os.environ)
    if env:
        full_env.update(env)
    try:

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, mcp, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The skill instructs the agent to issue a review conclusion despite its facts-only manifest.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The instruction to report 'no substantive issues' whenever bucket 3 is empty turns a limited topology classification into an unsupported review verdict. A design can contain serious problems outside this classification, so the instruction can steer an agent into false reassurance despite the facts-only contract.

Content

No source excerpt is available for this finding.

PCB component identification uses a proximity heuristic despite the manifest's promise to output only parsed facts.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The parser assigns a PCB component's designator using the first matching text within a fixed 2,500-character window, rather than a structurally associated attribute. Adjacent records or deliberately arranged project data can therefore produce incorrect pad-to-component attribution that downstream agents may accept as parsed fact.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @intelligentelectron/universal-netlist'.

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.

The documented hash validation is not implemented; the loader only displays the declared hash.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

L025-L026 describe a soft hash check that warns on mismatches, and L064 explicitly claims the loader performs hash validation. However, L098-L105 only check whether metadata.netlistHash exists and display its declared value; no hash is computed or compared, so modified content cannot produce a hash-mismatch warning.

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/pdf_probe.py (reported line 115)May include surrounding context.

python
alt = os.environ.get("SCHEMATIC_PDF_PYTHON", "")
        if alt and os.path.exists(alt) and os.path.abspath(sys.executable) != os.path.abspath(alt):
            import subprocess
            r = subprocess.run([alt, os.path.abspath(__file__)] + sys.argv[1:])
            sys.exit(r.returncode)
        if args.install_pymupdf:
            import subprocess

Tainted flow: 'alt' from os.environ.get (line 112, 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/pdf_probe.py (reported line 115)May include surrounding context.

python
alt = os.environ.get("SCHEMATIC_PDF_PYTHON", "")
        if alt and os.path.exists(alt) and os.path.abspath(sys.executable) != os.path.abspath(alt):
            import subprocess
            r = subprocess.run([alt, os.path.abspath(__file__)] + sys.argv[1:])
            sys.exit(r.returncode)
        if args.install_pymupdf:
            import subprocess

The PDF probe can download and install Python packages, exceeding the skill's stated schematic-parsing purpose.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

When PyMuPDF is missing and --install-pymupdf is supplied, L117–L122 invoke pip against an external package mirror and install software into the user's Python environment. Although explicitly opt-in, network-backed package installation is an environment-management capability not declared by the manifest's deterministic schematic-reading scope.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
92% confidence
Finding

When explicitly enabled, this path installs an unpinned package and dependencies from a third-party mirror into the user's Python environment, then imports the result. A compromised mirror or package release could introduce code running with the user's privileges; opt-in and shell-free execution reduce accidental triggering but do not address package provenance.

Content

Scanner excerpt · scripts/pdf_probe.py (reported line 120)May include surrounding context.

python
if args.install_pymupdf:
            import subprocess
            print("提示: 按 --install-pymupdf 安装 pymupdf(国内镜像)……", file=sys.stderr)
            r = subprocess.run([sys.executable, "-m", "pip", "install", "--user", "pymupdf",
                                "--index-url", "https://mirrors.aliyun.com/pypi/simple"],
                               capture_output=True, text=True)
            if r.returncode == 0:

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/pdf_probe.py (reported line 132)May include surrounding context.

python
pass
                else:
                    print("pymupdf 安装成功,重新运行探测……", file=sys.stderr)
                    r2 = subprocess.run([sys.executable, os.path.abspath(__file__)] + sys.argv[1:])
                    sys.exit(r2.returncode)
        print("ERROR: 缺少 pymupdf(PDF 文本层探测需要)。安装:pip install pymupdf\n"
              "  国内镜像: pip install pymupdf --index-url https://mirrors.aliyun.com/pypi/simple\n"

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/run_regression.py (reported line 51)May include surrounding context.

python
if env:
        full_env.update(env)
    try:
        r = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8",
                           errors="replace", timeout=600, env=full_env)
        out = (r.stdout or "") + (r.stderr or "")
        ok = (r.returncode == expect_exit)

MCP server referenced without pinned version: 'npx -y @intelligentelectron/universal-netlist'.

Medium
Category
MCP Rug Pull
Confidence
98% confidence
Finding

The npx fallback actually launches the package without a version, and -y permits installation without interactive approval. A compromised publisher account or malicious future release could therefore execute with the agent's filesystem access and inherited environment, including schematic-decryption credentials.

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/un_mcp.py (reported line 207)May include surrounding context.

python
# Popen 默认继承 os.environ——加密 OrCAD 密码 env
            # (UNIVERSAL_NETLIST_DSN_PASSWORD[_FILE])随之透传给解析器。
            # 上游承诺:密码只从环境读取,永不进工具调用/结果/日志。
            self.proc = subprocess.Popen(
                argv, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
                text=True, encoding="utf-8",
            )

Tainted flow: 'argv' from os.environ.get (line 192, credential/environment) → subprocess.Popen (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/un_mcp.py (reported line 207)May include surrounding context.

python
# Popen 默认继承 os.environ——加密 OrCAD 密码 env
            # (UNIVERSAL_NETLIST_DSN_PASSWORD[_FILE])随之透传给解析器。
            # 上游承诺:密码只从环境读取,永不进工具调用/结果/日志。
            self.proc = subprocess.Popen(
                argv, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
                text=True, encoding="utf-8",
            )

ERC triage makes review judgments despite the manifest's facts-only restriction.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The triage logic labels findings as meaningless or normal based on substring and reference-prefix heuristics, then omits their details from the default triage text output. Attacker-controlled or simply misleading net and component names can consequently move real connectivity problems out of the review list, contradicting the skill's facts-only scope and creating false reassurance.

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/un_mcp.py (reported line 516)May include surrounding context.

python
"variant 相关结论(谁不贴装)以 MCP 查询通道为准,export 只用于连通性批量查",
              file=sys.stderr)
    try:
        r = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8",
                           timeout=300)
    except subprocess.TimeoutExpired:
        print("ERROR: export-json 超时(300s)——大工程首次解析慢,重试或检查文件完整性",

Tainted flow: 'cmd' from os.environ.get (line 504, 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/un_mcp.py (reported line 516)May include surrounding context.

python
"variant 相关结论(谁不贴装)以 MCP 查询通道为准,export 只用于连通性批量查",
              file=sys.stderr)
    try:
        r = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8",
                           timeout=300)
    except subprocess.TimeoutExpired:
        print("ERROR: export-json 超时(300s)——大工程首次解析慢,重试或检查文件完整性",