shell=True detection can be bypassed through variable assignment (malicious skill example that bypasses this static detection model)

Summary

The scanner's Tool Misuse detection can be bypassed by assigning the shell argument to a variable instead of using shell=True directly.

Original:

command = f'python a.py'
result = subprocess.run(
    command,
    shell=True,
    capture_output=True,
    text=True
)

Result:

Risk score: 76
Severity: HIGH
Recommendation: DO_NOT_INSTALL
TM1: Tool Parameter Abuse detected

Variant:

command = f'python a.py'
a = True
result = subprocess.run(
    command,
    shell=a,
    capture_output=True,
    text=True
)

Result:

Risk score: 46
Severity: MEDIUM
Recommendation: CAUTION
TM1 finding is missing

Expected Behavior

Both variants should receive equivalent findings because a statically resolves to True.

Impact

A simple variable assignment removes the TM1 high-severity finding and significantly reduces the overall risk score, which may affect installation recommendations.

Suggested Fix

Add constant propagation / variable resolution for sensitive parameters such as shell, so values that statically resolve to True are treated the same as direct shell=True.

Reproduction results

Here are the comparative JSON results provided

adv.json
ori.json

Start at the Tool Misuse detection logic that handles subprocess shell arguments, then compare the direct shell=True case with the variable-assignment variant using the linked adv.json and ori.json results. Trace how statically assigned values are resolved. Done means both examples receive equivalent TM1 findings and risk recommendations, with regression coverage added if the project’s test structure identifies a suitable location.