Skill Orchestra

• Outcome: Pass
• Latest audit: Sep 11, 2026 · 1:34 AM UTC
• Version: 2.0.0

Overview

This skill is a disclosed routing helper with some overclaimed or unfinished features, but the inspected artifacts do not show hidden data access, persistence, exfiltration, or destructive behavior.

Install only if you are comfortable with a routing helper that may keep or learn from routing decisions if its documented v2 examples are completed later. The current bundled runtime looks low-risk, but its wrapper and documentation are incomplete, so expect possible functional issues rather than security impact.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (7)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 93% confidence
• Finding: The declared description promises a specific routing system with explicit competence/cost modeling and performance characteristics. The supplied code does not implement those behaviors; instead, it is a generic adapter/wrapper around an external 'skill_orchestra.py' implementation. Its only visible routing-related logic is simple keyword matching in can_handle(). It also dynamically imports and executes another class, which is a broader and more generic capability than the declared purpose. Because the primary visible behavior is wrapper/delegation rather than the claimed modeled router, the description does not accurately represent what this code chunk itself actually does.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The regex patterns are extremely broad and match common verbs like 'analyze', 'write', 'create', 'find', and 'solve', which can cause the router to activate on a large share of ordinary prompts. Overbroad trigger scope can route unrelated or sensitive user requests through this skill unexpectedly, increasing the chance of unnecessary context capture, unintended agent selection, or downstream processing not expected by the user.

Description-Behavior Mismatch

Medium

• Confidence: 85% confidence
• Finding: The manifest describes a routing skill focused on competence/cost-based agent selection. The v2.0.0 documentation introduces a pattern learner that loads routing history from a file ("memory/routing-patterns.json"), expanding the skill into persistent learning/behavior-tracking rather than just performing routing decisions.

Description-Behavior Mismatch

Medium

• Confidence: 88% confidence
• Finding: The manifest claims skill-aware agent routing for current requests, but the predictive router goes further by anticipating likely next contexts and pre-routing them in advance. That proactive speculative behavior is not clearly conveyed by the manifest's limited routing description.

Description-Behavior Mismatch

Medium

• Confidence: 92% confidence
• Finding: The manifest describes explicit competence/cost modeling for agent routing, but this wrapper only registers a generic capability and uses simple substring checks in can_handle before blindly dispatching to whichever method exists on a dynamically loaded class. That behavior does not implement the described routing model and materially understates the gap between the claimed orchestration logic and the actual code in this file.

Dynamic attribute access via getattr()

Low

• Category: Dangerous Code Execution

Content

# Find the main class
            for name in dir(module):
                obj = getattr(module, name)
                if isinstance(obj, type) and name != "UnifiedSkill":
                    self._original = obj()
                    break

• Confidence: 50% confidence
• Finding: Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Intent-Code Divergence

Low

• Confidence: 87% confidence
• Finding: The method is documented as registering the skill's capabilities, yet the inline TODO and single "general" registration show it is not actually describing the skill's real routing/orchestration competence. This is a direct intent/documentation mismatch rather than mere omission because the code presents a placeholder as if it were the implemented capability set.

Static analysis

No suspicious patterns detected.