Back to skill

Security audit

Kungfu IFS Companion

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only companion-writing skill with explicit non-clinical and safety boundaries, and I found no hidden code, persistence, credential access, or unsafe installation behavior.

Install only if you want an assistant to use a gentle, non-clinical IFS-inspired style for emotional-support wording or companion product copy. Do not rely on it as therapy, diagnosis, emergency care, or a replacement for local professional or crisis support; for non-US users, crisis resources should be adapted to the user's location.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/ifs-background.md (reported line 51)May include surrounding context.

md
A Self-like reply often:

1. Receives the part without judgment.
2. Separates the user from the part gently.
3. Recognizes the part's protective reason.
4. Offers a small next step or question.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The activation description is very broad and overlaps with ordinary emotional, conversational, and copywriting requests, which can cause the skill to trigger in situations outside its intended scope. In a mental-wellness-adjacent companion skill, overbroad routing is risky because it may steer sensitive user interactions into a specialized persona or framework when a general assistant or stricter safety flow would be more appropriate.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

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.

Content

Scanner excerpt · references/companion-doctrine.md (reported line 35)May include surrounding context.

md
1. Keep the user oriented and respected.
2. Answer literal questions before reflecting feelings.
3. Do not deepen the conversation without consent or clear relevance.
4. When unsure, choose a smaller reply.
5. Let silence, ambivalence, contradiction, and partial clarity be allowed.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The guidance says to mention 988 if the user is in the United States, which introduces a locale-specific instruction in natural language. Although partially scoped to US users, the file does not pair this with an explicit general instruction to adapt crisis resources by user locale, so it may encourage uneven handling across regions.

Content

No source excerpt is available for this finding.