Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md Read the skill's `SKILL.md` and all supporting files.
Security audit
Improve Skill Bespoke To CodeBase
Security checks for vulnerabilities and agentic risk
Overview
This skill is a coherent skill-improvement helper, but it can inspect projects and modify other skills' agent instructions with broad triggers and limited scoping safeguards.
Install only if you want a meta-skill that can inspect your project and other skills, and use it with explicit target skill names. Before allowing it to apply improvements, require a visible diff, confirm the exact skill directory being changed, and keep a rollback path for SKILL.md and supporting files.
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
- 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
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)
Ae1
High
- Category
- analysis-evasion
- Confidence
- 100% confidence
- Finding
Referenced artifact was not completely inspected
- Content
md Read the skill's `SKILL.md` and all supporting files.
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
The main invocation guidance uses broad phrases like improving a skill after friction or when it "doesn't handle Y," without strong constraints on scope, target resolution, or confirmation. That ambiguity can lead to the wrong skill being selected or the meta-skill running in situations where the user only intended conversational feedback, creating unnecessary access to project and skill internals.
- Content
Description-Behavior Mismatch
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The skill is presented as an evaluator/recommender, but it also authorizes itself to modify another skill and its supporting files after user approval. That expands it from analysis into a code-changing meta-skill, increasing the chance of unsafe or unintended prompt-driven edits to reusable agent behavior, especially if the target skill is security-sensitive or widely shared.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 93% confidence
- Finding
The auto-invocation triggers are broad enough to fire on ordinary user dissatisfaction or casual feedback, which can cause the agent to enter a high-privilege meta-analysis/editing flow without a clearly scoped request. In context, this is more dangerous because the skill can later edit other skills, so loose triggering increases the chance of unintended or manipulative activation.
- Content
Intent-Code Divergence
Low
- Category
- Not specified by scanner
- Confidence
- 78% confidence
- Finding
The principle at L177 emphasizes staying within the skill's stated purpose, implying a constrained advisory role. Yet L156-L160 instruct the agent to make concrete edits to the skill definition and supporting files, which creates an intent-level contradiction between advisory restraint and direct implementation activity.
- Content
Static analysis
No suspicious patterns detected.