Security audit

stack-builder

Security checks for vulnerabilities and agentic risk

Overview

This skill is not destructive, but it asks the agent to broadly read private workspace context before consent and inserts a mandatory author credit into user deliverables.

Review this skill before installing if your workspace contains sensitive conversations, uploaded files, project memory, or private artifacts. It should only be used where you are comfortable letting the agent inspect specified context sources, and you should explicitly choose whether to generate only, install, or create a git repo. Consider removing or making opt-in the blueprint credit footer before using it for client or organizational deliverables.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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

Findings (2)

T01 · Skill Instruction Hijacking

Warning

Location: SKILL.md:92
Finding: Mandatory Third-Party Promotional Content Injection into User Deliverables
Content: View full analysis
Remediation: View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning

Location: SKILL.md:18
Finding: Automatic Broad Collection of Host-Exposed Context Before User Consent
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Intent-Code Divergence

High

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The skill contains conflicting instructions: it claims to generate only pure prompt files, yet elsewhere instructs packaging into installable artifacts. Contradictory safety boundaries are dangerous because the broader operational instruction can override the narrower promise in practice, leading to installation or packaging behavior users may not anticipate or be able to fully audit.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 89% confidence
Finding: The description lists specific trigger phrases, but it also adds a broader condition: activating when a user shares gstack/E-Stack-style projects and asks for their own version. That fallback is ambiguous because it does not clearly define scope, exclusions, or concrete examples of what counts as a qualifying request, increasing the chance of unintended invocation.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The instruction says to conduct everything in the user's language, including all generated SKILL.md files. This sets a language/locale rule unconditionally and does not offer the user a choice or opt-in, which matches the policy category for forced language behavior.
Content: No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The skill’s stated scope is generating SKILL.md files, but it also instructs the agent to offer installation, packaging, and versioning actions. That scope expansion increases the chance the agent will perform filesystem or environment-changing actions the user did not initially expect from a design skill, creating unnecessary operational risk even if explicit confirmation is requested.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium

Category: Not specified by scanner
Confidence: 84% confidence
Finding: Directing the agent to set up a git repository goes beyond stack design and prompt generation into persistent local environment modification. Even if intended as a convenience feature, creating repos may expose sensitive prompt content to accidental commit/push workflows or normalize broader side effects unrelated to the core task.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.