← back to prompt-injection-defense

SkillSpector · prompt-injection-defense

independent scanner by NVIDIA · skill by seb1n · how it works ↗

CAUTIONmax severity: MEDIUMrisk score: 23

Without declared permissions the skill's intent is opaque and cannot be validated.; Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, dat…; Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous …

scanned 2026-08-23

Findings (4)

MEDIUMMCP Least Privilegeconfidence: 0.7

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

SKILL.md

MEDIUMExcessive Agencyconfidence: 0.75

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

scripts/audit_boundary_manifest.py

MEDIUMExcessive Agencyconfidence: 0.75

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

scripts/audit_boundary_manifest.py

HIGHTool Misuseconfidence: 0.255

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

references/defense-patterns.md

What the verdicts mean

SkillSpector reports on SkillMD's shared five-tier scale. See how SkillSpector works ↗.

PASS

Overall severity LOW (risk score in the safe range)

CAUTIONthis skill

Overall severity MEDIUM

WARNING

Overall severity HIGH

FAIL

Overall severity CRITICAL

INCONCLUSIVE

Scan could not complete