Back to skill

Security audit

lgd-evo

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign with medium confidence: scanners report a dispatcher whose listed behavior fits its stated purpose, alongside a broken advertised availability check and several permission or trigger concerns; one referenced path is missing, so evidence is incomplete.

The verdict is benign, but it is not a guarantee of safety, and the scan was partial because a referenced path was missing. Before installing, consider whether the broad trigger and declared environment permissions fit your needs; the advertised availability check is reported as nonfunctional.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (6)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The listing and guidance behavior matches the declared thin-dispatcher purpose, and there is no evident unrelated or covert capability. Nevertheless, availability probing is an explicitly advertised core feature that the supplied CLI cannot perform. Although verify() contains directory checks and an optional child-script --help probe, --verify crashes before calling it. The optional JSON report is a supporting implementation detail, not an independent undeclared capability.

Content

No source excerpt is available for this finding.

Code capability 'env' detected in scripts/lgd_evo.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.

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/lgd_evo.py (reported line 70)May include surrounding context.

python
note = "发现脚本;默认不执行(加 --execute 才实跑)"
                else:
                    try:
                        r = subprocess.run([sys.executable, script, "--help"],
                                           capture_output=True, text=True, timeout=25)
                        rc = r.returncode
                        status = "LIVE" if rc == 0 else "ERROR"

Tainted flow: 'script' from os.environ.get (line 58, 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/lgd_evo.py (reported line 70)May include surrounding context.

python
note = "发现脚本;默认不执行(加 --execute 才实跑)"
                else:
                    try:
                        r = subprocess.run([sys.executable, script, "--help"],
                                           capture_output=True, text=True, timeout=25)
                        rc = r.returncode
                        status = "LIVE" if rc == 0 else "ERROR"

The advertised one-command availability probe always fails: --verify accesses args.execute, but the argument parser never defines --execute.

Medium
Category
Excessive Agency
Confidence
100% confidence
Finding

The manifest promises a one-command probing entry point for the evolution subskill. However, the parser at L085–L092 never creates an execute attribute, so every --verify invocation raises AttributeError before probing any node or producing evidence; supplying --execute instead fails argument parsing.

Content

No source excerpt is available for this finding.

The standalone trigger `演化` (“evolution”) is too broad for an AI-agent evolution dispatcher.

Low
Category
Trigger Abuse
Confidence
87% confidence
Finding

The trigger list includes the generic word 演化, which can occur in discussions of biological evolution or other unrelated changes. Although the surrounding description explains the skill's purpose, it does not explicitly constrain this trigger to AI-agent self-improvement requests or exclude unrelated uses.

Content

No source excerpt is available for this finding.