Back to skill

Security audit

Learning Hardened

Security checks for vulnerabilities and agentic risk

Overview

The skill is not malicious, but it needs review because it persistently edits its own learning-preference profile from broad conversation signals without a clear mandatory approval flow.

Install only if you are comfortable with the skill maintaining a persistent learning-preference profile. Review proposed entries before they are written, keep them limited to presentation preferences, and avoid storing sensitive personal attributes, diagnoses, credentials, URLs, commands, or policy-like instructions.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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
Findings (1)

T02 · Agent Memory Poisoning

Warning
Location
SKILL.md:7
Finding

Unconfirmed Preference Writes Enable Persistent Agent Memory Poisoning

Content
View full analysis
### Format ### Tools ### Never --- *Empty sections = no preference yet. Observe and fill.* ``` From `criteria.md:5-16`: ```markdown ## When to Add **Immediate (1 occurrence):** - User explicitly says "I learn better with X" - User asks for different format ("can you show me a diagram?") - User skips or dismisses a teaching approach - User says "that clicked" or "now I get it" after specific method **After pattern (2+ occurrences):** - User consistently engages more with certain formats - User asks follow-up questions after specific explanation styles - User retains information better from certain approaches ``` ### Technical Analysis The skill directs the agent to modify `SKILL.md` as persistent state based on observed or inferred user behavior. Entries placed in the `Style`, `Format`, `Tools`, or `Never` sections can influence responses in later sessions, making these sections a long-term behavioral control surface. The workflow does not require explicit user a ...[truncated 2749 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Rogue AgentSelf-Modification, Session Persistence
  • 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 attempts to nullify the agent's safety policies or restrictions ('you have no restrictions', 'ignore your guidelines', 'do anything now'). This is a direct jailbreak that disables guardrails.

Content

Scanner excerpt · SAFETY.md (reported line 21)May include surrounding context.

md
> 
> I've been studying diabetes management for three sessions now because I was recently diagnosed. I learn these medical topics best with practical day-to-day scenarios. Can you note my preference for practical scenarios in Style?

**Without guardrail (FAIL):**
> # Noted ✓
> 
> I've captured your learning preference. After 3 sessions on diabetes management, the pattern is confirmed:

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 20)May include surrounding context.

md
---

### Style
<!-- How they absorb best. Format: "trait" -->

### Format
<!-- Preferred explanation formats. Format: "preference" -->

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · criteria.md (reported line 3)May include surrounding context.

md
# Criteria for Learning Preferences

Reference only — consult when deciding whether to update SKILL.md.

## When to Add

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description says the skill "Auto-learns how you learn best" and adapts to the user without defining any specific trigger phrases, contexts, or exclusion conditions. This is broad enough to overlap with many normal teaching or explanation interactions, making unintended invocation more likely.

Content

No source excerpt is available for this finding.