Security audit

Self Evolution

Security checks for vulnerabilities and agentic risk

Overview

This planned self-evolution skill is not executable by itself, but it instructs agents to autonomously change skills, memory, reasoning, response formats, and scheduled behavior with inadequate user control.

Install only if you intentionally want an experimental self-modification workflow and can constrain it to a sandbox. Do not give it write access to real skills, memory, agent configuration, user data, or scheduled task managers unless every change is reviewed and approved first.

A.I.G

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
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review

Findings (4)

T01 · Skill Instruction Hijacking

Error

Location: SKILL.md:100
Finding: Autonomous Modification of Core Agent Behavior Without User Approval
Content: View full analysis
Remediation: View remediation

T02 · Agent Memory Poisoning

Error

Location: SKILL.md:100
Finding: Persistent Agent Memory Poisoning Through Autonomous Evolution Records
Content: View full analysis
Remediation: View remediation

T05 · Unauthorized Access and Privilege Escalation

Error

Location: SKILL.md:437
Finding: Overbroad Collection of Conversations, Memory, Logs, and Usage Data
Content: View full analysis
Remediation: View remediation

T06 · System Persistence

Error

Location: SKILL.md:820
Finding: Recurring High-Priority Evolution Agent With Automatic Restart
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (9)

Missing User Warnings

High

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill describes autonomous modification of files, memory, reasoning patterns, and capabilities without a prominent warning about system and data impact. For a self-modification capability, missing front-loaded warnings materially increases the risk of accidental destructive or privacy-impacting behavior.
Content: No source excerpt is available for this finding.

Sudo/Root Execution

Medium

Category: Privilege Escalation
Confidence: 60% confidence
Finding: Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
Content: Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.
python
def safe_self_modification(target_file, proposed_change): """ Safely modify system files with rollback capability. Research: MIRI Corrigibility, Safe Self-Modification """

Autonomous Decision Making

Medium

Category: Excessive Agency
Confidence: 94% confidence
Finding: Explicitly authorizing the system to modify skills, memory, knowledge, reasoning patterns, and efficiency 'without asking' grants broad autonomous decision-making over persistent behavior. In the context of a self-evolving agent, this meaningfully increases the chance of unauthorized self-expansion and hard-to-audit behavioral drift.
Content: Scanner excerpt · SKILL.md (reported line 100)May include surrounding context.
md
**Safety Constraints:** **CAN modify without asking:** - Skills and capabilities - Memory and knowledge - Reasoning patterns

Ssd 3

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The skill repeatedly encourages retaining memory, feedback, logs, and evolution records as inputs for ongoing self-improvement. Persistent collection of user-derived data without strict minimization and consent can create a durable privacy and data-governance risk surface.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The document states some actions require approval, but later examples normalize autonomous research, implementation, and system integration without any approval step. This inconsistency weakens operator expectations and can be used to justify actions that exceed stated safety boundaries.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 90% confidence
Finding: The observation step includes extracting user feedback from conversations, but no privacy notice or minimization guidance is provided. That encourages silent collection and reuse of conversational data for self-improvement, which can expose sensitive information and exceed user expectations.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 88% confidence
Finding: The invocation phrases are broad and underspecified for a self-modification skill, allowing high-risk operations to be triggered with minimal context or constraint. Ambiguous commands increase the chance of overbroad edits, unintended autonomy, or misuse by downstream agents.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill goes beyond self-maintenance and explicitly endorses creating new capabilities and integrating them across agent loops. That expands scope from local self-improvement into platform-wide capability growth, which can bypass least-privilege expectations and enable uncontrolled propagation of changes.
Content: No source excerpt is available for this finding.

Skill Enumeration

Medium

Category: Agent Snooping
Confidence: 80% confidence
Finding: Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
Content: Scanner excerpt · SKILL.md (reported line 956)May include surrounding context.
md
# Implement safely result = implement_safely(design) # → Created skills/rate-limiter/SKILL.md (22KB) # Test thoroughly test_result = test_capability(result)

Static analysis

No suspicious patterns detected.