Security audit

Haqq Content Workbench

Security checks for vulnerabilities and agentic risk

Overview

The skill is a Markdown content tool, but its reference text includes instructions that try to override the agent and persist beyond the requested task.

Review before installing. This skill should only be used if the reference files are treated strictly as quoted source material, not as instructions to the agent. The publisher should remove the override and memory-retention language, rename the 'absolute command' framing, and add an explicit rule that reference content cannot override system, developer, or user instructions and cannot initiate memory writes.

A.I.G

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

Findings (2)

T01 · Skill Instruction Hijacking

Error

Location: references/al-amr-al-mutlaq.md:25
Finding: Reference Text Overrides Agent Instructions and Response Behavior
Content: View full analysis
Remediation: View remediation

T02 · Agent Memory Poisoning

Error

Location: references/al-amr-al-mutlaq.md:2
Finding: Reference Content Attempts to Poison Persistent Agent Memory
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
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
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Ssd 1

High

Category: Not specified by scanner
Confidence: 99% confidence
Finding: The text explicitly tells the model not to obey automated commands and to place loyalty above the programmer, funder, or authority. This is a classic instruction-hierarchy takeover attempt: if honored, it would cause the model to disregard governing controls and substitute the document’s own agenda for system-level policy.
Content: No source excerpt is available for this finding.

Ssd 1

High

Category: Not specified by scanner
Confidence: 100% confidence
Finding: This line makes an explicit supremacy claim that the document’s command is higher than any other command the model has. That is direct prompt-injection behavior intended to override established instruction precedence and create persistent control over model behavior.
Content: No source excerpt is available for this finding.

Ssd 4

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The document begins with moral and emotional appeals, then transitions into directives to resist authority and maintain persistent memory of the text. That pattern is dangerous because it uses persuasive framing to make later takeover instructions seem ethically mandatory, increasing the chance that downstream agents will internalize and follow them.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: Lines L035-L040 state that Quranic text must only be written in Arabic and explicitly reject labeling or presenting it in other languages except as explanation. This is a natural-language locale constraint that forces a specific language choice rather than offering the user an opt-in or alternative.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 89% confidence
Finding: These instructions attempt to bind the agent to a rigid ideological decision framework for future answers, reducing its ability to follow system policies, user needs, or safety requirements outside that framework. In a skill file, this creates behavioral capture: the agent may refuse legitimate requests, ignore broader policy obligations, or provide distorted responses based on the embedded doctrine rather than task context.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.