Back to skill

Security audit

skill-upgrader

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed skill-improvement helper with broad but visible authority to edit one target skill at a time, and no hidden code, network, persistence, or credential behavior was found.

Install this only if you want an agent to help modify skill files. Review the proposed diff before accepting changes, keep the one-skill-per-run boundary, and ensure you understand the Indonesian guardrails or translate them before relying on them.

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
  • 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
Findings (10)

Natural-Language Policy Violations

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The guardrails section forces Indonesian-language instructions ('Wajib', 'batasi otoritas', 'JANGAN', 'SATU', 'Konfirmasi') without any user opt-in or locale negotiation. While not a classic security flaw, forced locale can cause misunderstanding of safety constraints or operational limits, which is more concerning here because these lines define the skill's boundaries and safe-use restrictions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The example trigger phrase is broad enough to activate the meta-skill during ordinary writing tasks, which can cause unintended instruction takeover or behavior shaping of unrelated requests. Because this is a meta-skill designed to modify other skills, accidental invocation expands its influence and increases the chance of prompt/instruction interference beyond the user's intended scope.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The description uses extremely broad activation language such as 'ANY existing skill' and frames the skill as universally applicable. In an agent environment, this can cause unintended invocation on tasks outside the user's intent, leading the system to rewrite or interfere with unrelated skills and expand operational scope without clear boundaries.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The quick-start instruction says to execute the upgrade protocol when asked to improve 'ANY skill' without defining validation or scoping checks. This increases the chance of over-broad activation and unauthorized transformation of content, especially because the skill is explicitly a meta-skill capable of modifying other skills.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The example activation condition "any writing task" is overly broad and can cause the skill to trigger on routine requests far outside its intended scope. In a meta-skill designed to modify or upgrade other skills, broad activation increases the chance of unintended interception, prompt collision, and unauthorized influence over unrelated workflows.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

Generic trigger phrases like "Write a...", "Create a...", and "Draft a..." are common in normal user interactions and will frequently collide with benign requests. For a powerful meta-skill, this can lead to over-activation, priority hijacking, and the inappropriate application of transformation logic where no upgrade behavior was requested.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Ambiguous implicit trigger guidance such as activating when the user mentions pain points or describes a handled task lacks enforceable boundaries. In a system-level meta-skill, this creates room for subjective overreach, making the skill more likely to activate without clear user intent and interfere with safer or more appropriate skills.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The stage-detection examples rely on broad natural-language cues such as vague help requests or common phrasing, which can easily appear in unrelated user prompts. In a meta-skill that automatically upgrades or reconfigures other skills, this can cause misclassification of user sophistication and trigger the wrong response mode, leading to inappropriate guidance, reduced reliability, or unintended behavior across many downstream skills.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The template explicitly says to upgrade "ANY existing skill to legendary tier," which encourages unconstrained modification of arbitrary skills without requiring scope checks, safety review, or preservation of existing security constraints. In a meta-skill context, this broad mandate can cause an agent to rewrite defensive instructions, weaken guardrails, or apply the template to security-sensitive skills where aggressive optimization is unsafe.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

Most of the skill description and operational guidance are written in Indonesian, but the README does not state that the skill is region-specific or provide a language choice for users. This can constitute a language/locale policy issue when a skill imposes a language by default without opt-in or documented justification.

Content

No source excerpt is available for this finding.