Back to skill

Security audit

Cybernetic Evolver

Security checks for vulnerabilities and agentic risk

Overview

This skill is not overtly malicious, but it asks users to wire a persistent self-optimization loop into an agent using broad hooks and an unsafe workspace import pattern.

Review before installing. Use this only in an isolated workspace, avoid adding writable directories to the front of Python's import path, provide and verify any init_evolver.py loader yourself, restrict save/load paths to a dedicated evolver state directory, and require explicit user approval before its recommendations change agent routing or strategy.

Vulnerability Patterns
  • Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T07 · Tool Hijacking and Spoofing

Error
Location
SKILL.md:445
Finding

Workspace Import-Path Precedence Enables Python Module Hijacking

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
DEMO/example.py:124
Finding

Demonstration Script Performs Hard-Coded Writes Outside the Project Directory

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (21)

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document title and all descriptive prose are written in Chinese, with no indication that users may choose another language or that the skill is region-specific. Under the policy, forcing a specific language without opt-in is a natural-language locale violation.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring and user-facing documentation are written entirely in Chinese, with no indication that language is selectable or that the skill is intentionally region-specific. This creates a natural-language locale policy issue because the skill effectively imposes a specific language on users without opt-in.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill provides generic hooks for external state extraction, action transformation, metric computation, and an optimize loop that can steer arbitrary agent actions. In a broader agent framework, these hooks can be wired to sensitive capabilities so the evolver effectively becomes a policy engine for choosing external operations beyond its stated purpose, increasing the risk of abuse, privilege creep, and hard-to-audit behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The module exposes unrestricted save/load primitives that read from and write to arbitrary file paths supplied by the caller. In an agent setting, this expands the skill from self-evolution logic into filesystem access, enabling unintended overwrite of local files, loading attacker-controlled state, or persistence of sensitive optimization history if the surrounding agent passes untrusted paths.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This Python file contains user-facing documentation and runtime messages in Chinese, beginning with the module docstring, without any indication that language selection is optional. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy violation even when it appears in code strings or comments.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill documentation is written entirely in Chinese, including the core description, headings, and usage explanations, with no indication that users may choose another language. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is justified, which is not present here.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill describes persistent state save/load operations and workspace file access, but declares no explicit tool scope or permission boundaries. In an agent ecosystem, undocumented file read/write capability can enable unintended persistence, modification of local state, or broader workspace interaction without operator awareness.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest trigger list contains generic phrases without contextual constraints, making invocation collisions likely in normal conversations about adaptation or self-improvement. In this skill's context, activation can influence agent strategy selection and persistence behavior, so loose triggers create a real control-surface risk even without overtly malicious code.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger phrases include broad terms such as '自我进化', 'adaptive AI', and 'self-improving', which can match ordinary discussion and cause the skill to activate outside intended contexts. Because this skill proposes optimization, delegation, persistence, and structural mutation behaviors, accidental activation increases the chance of unexpected autonomous changes or recommendations in unrelated conversations.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file is written entirely in Chinese, including headings, workflow steps, and operational guidance, with no indication that language choice is optional or that the skill is region-specific. Under the stated policy, forcing a specific language without user opt-in is a natural-language locale violation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The optimize() docstring describes a control loop in which the evolver selects an action, the external system executes it, and a result is then obtained. In code, no execution callback is invoked and no new result is fetched; instead, the method repeatedly reads a preexisting last_result from the input context, which contradicts the documented closed-loop behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The manifest describes a closed-loop adaptive/self-organizing AI framework, but this demo script performs concrete filesystem output by saving plots to fixed workspace paths. While plotting can support a demo, persistent file generation is still behavior beyond the core self-evolution functionality described in the manifest.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

These lines save demo-generated visualizations to fixed paths, which is not reflected in the manifest description focused on adaptive control, feedback loops, mutation, and online learning. The behavior is not harmful by itself, but it is an externally visible capability not described at the manifest level.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The manifest presents the skill as an AI self-evolution/control framework, but the script also ensures directories exist and emits image artifacts into the workspace. That write-oriented demo tooling is a broader operational behavior than the conceptual framework description suggests.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The primary natural-language usage instructions and trigger guidance are presented in Chinese, including the explicit activation wording, without stating that users may choose another language. This can be a language/locale policy issue when a skill effectively assumes one language by default without opt-in or justification.

Content

No source excerpt is available for this finding.