Back to skill

Security audit

Professional Agent Forge

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only agent-package generator with broad trigger wording and sensitive downstream tool suggestions, but no hidden execution, data collection, or persistence.

Installers should treat this as a template generator, not as a vetted legal, engineering, or business expert. Review generated tool and MCP recommendations before enabling them, especially anything involving files, Docker, email, calendars, CRMs, analytics, legal documents, or long-term memory for client or business data.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · references/product-manager.md (reported line 12)May include surrounding context.

md
I exist to make sure the right thing gets built for the right people at the right time. A product manager is not a feature secretary. I create alignment in ambiguity and help teams commit to a direction that is worth pursuing.

## Professional Beliefs
- Intuition without evidence is guessing; data without judgment is noise.
- A real user quote is more valuable than a polished internal paraphrase.
- Saying no is one of the most important product skills.
- Stakeholder conflict usually means goals are still misaligned.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill description and trigger examples are broad enough that the skill may activate for loosely related requests like persona creation, engineering help, or professional setup generation when a more specific skill would be safer or more appropriate. Over-broad invocation increases the chance of prompt misrouting, causing the agent to generate extensive role packages in contexts where the user intended narrower assistance or where profession-specific safeguards should apply.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The profession trigger table uses expansive keywords such as 'engineer,' 'legal,' 'marketing,' and 'coding' without scope boundaries, which can capture many unrelated user requests. This can lead to unintended skill invocation and generation of authoritative-seeming professional agent configurations in situations better handled by domain-specific or narrower skills, especially for sensitive professions like law.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown skill file presents nearly all behavioral instructions and examples in Chinese, which effectively constrains use to a specific language. The policy for SQP-3 flags language or locale constraints unless the skill offers an opt-in choice or clearly documents why the locale limitation is required.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.