Fiverr

• Outcome: Review
• Latest audit: Sep 10, 2026 · 7:39 PM UTC
• Version: 1.0.0

Overview

This is a non-executable Fiverr business-advice skill, but parts of it teach users how to make automation look less detectable and give ambiguous account-scaling advice.

Review this skill carefully before installing. It is not malware and does not execute code, but users should avoid following the parts that describe what Fiverr detects, how to make responses look natural, or how to operate multiple accounts unless they have confirmed current Fiverr policy and written authorization where needed.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

Findings (2)

other

Warning

• Location: scaling.md:37
• Finding: Guidance Facilitates Platform Automation-Detection Evasion## Vulnerability Details **File Location**: `scaling.md:37-40, 83-93` **Vulnerability Type**: Platform automation-detection evasion **Risk Level**: Medium **Complete Relevant Snippet**: ```markdown ### Response Timing Rules - **Don't respond instantly** every time (looks automated) - Vary between 5-30 minutes during "work hours" - Occasional 1-2 hour delays are natural - Set "away" status during off hours ``` ```markdown ## Automation Safety ### What Fiverr Detects - Identical message text sent repeatedly - Response times consistently under 30 seconds - Login from multiple IPs simultaneously - Unusual activity patterns ### Safe Practices - Use template variations, never exact copies - Add natural delays (5-30 min typical) - One IP/device per account - Human final approval on all sends - Never automate: deliveries, cancellations, disputes ``` ### Technical Analysis The Skill enumerates signals reportedly used by Fiverr to detect automation and then recommends changing message text, response timing, and connection patterns to make activity appear natural. Although human approval is also recommended, the rationale explicitly connects these behavioral modifications to avoiding an automated appearance. This is not an executable software vulnerability and does not map to classifications T01–T09. It is unsafe operational guidance that could help users conceal automation-like activity from platform integrity controls. ### Attack Path 1. A user loads the Skill and follows its scaling workflow. 2. The user identifies the documented platform detection signals. 3. Automated or heavily templated messages are configured with textual variations. 4. Sends are delayed by 5–30 minutes rather than occurring immediately. 5. Account access is constrained to one apparent IP or device. 6. The resulting activity is intended to appear more human and become less detectable by platform controls. ### Impact A ...[truncated 436 chars]
• Remediation: ## Remediation Suggestions 1. Remove statements identifying platform detection signals and instructions to appear “natural.” 2. Replace detection-oriented advice with an explicit requirement that all automation comply with Fiverr's current terms and approved interfaces. 3. Recommend timing variation only for legitimate workload management, not concealment. 4. Require meaningful human review and customization based on each buyer's request. 5. Prohibit techniques intended to disguise automated platform access or messaging. 6. Advise users to obtain written platform authorization before deploying automation at scale. 7. Add a compliance note requiring periodic review of Fiverr's current automation policies because platform rules may change.

other

Note

• Location: scaling.md:103
• Finding: Contradictory Multiple-Account Guidance Could Encourage Prohibited Account Operation## Vulnerability Details **File Location**: `scaling.md:103` and `red-flags.md:54` **Vulnerability Type**: Ambiguous platform account-management guidance **Risk Level**: Low **Complete Relevant Snippets**: ```markdown | 200+ | Growth | Multiple accounts (legitimately), upsells | ``` ```markdown | Multiple accounts | Both accounts banned | ``` ### Technical Analysis The Skill gives conflicting advice. One document warns that multiple accounts result in both accounts being banned, while another recommends “Multiple accounts (legitimately)” as a growth action without defining an approved account type, authorization process, or applicable exception. This ambiguity could be interpreted as permission to create or operate additional seller accounts. The issue does not confer technical privileges and does not map to classifications T01–T09, but it may cause unsafe platform-account practices. ### Attack Path 1. A user reaches the scaling milestone for more than 200 monthly orders. 2. The user interprets “Multiple accounts (legitimately)” as authorization to create additional accounts. 3. Additional accounts are created or operated without confirming an applicable platform exception. 4. Fiverr correlates the accounts through identity, payment, device, network, or behavioral signals. 5. The accounts may be restricted or banned under the rule described elsewhere in the Skill. ### Impact Assessment No system privileges or unauthorized technical access are directly obtained. The affected scope is the user's Fiverr presence and associated business operations. Potential consequences include linked-account suspension, loss of marketplace access, interrupted orders, delayed funds, and reputational harm.
• Remediation: ## Remediation Suggestions 1. Remove the multiple-account recommendation unless a precise, currently valid Fiverr policy exception can be cited. 2. State that users must not create or operate additional accounts without prior written platform authorization. 3. Explain any approved team, agency, or business-account mechanism separately from ordinary seller accounts. 4. Add a cross-reference to the prohibition in `red-flags.md` to prevent contradictory interpretation. 5. Require users to verify current platform terms before changing account structure.

SkillSpector

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Context-Inappropriate Capability

High

• Confidence: 98% confidence
• Finding: The 'Automation Safety' section teaches users what Fiverr detects and how to structure behavior to evade those signals, including timing variation and device/IP practices. That goes beyond productivity advice and materially assists users in concealing potentially prohibited automation or coordinated account activity, increasing the risk of platform abuse.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: By describing platform detection signals and pairing them with 'safe practices' to avoid detection, the section provides actionable concealment guidance without warning about policy breach or enforcement consequences. In this business-scaling context, that makes the skill more dangerous because it can be directly used to operationalize stealthy non-compliant automation.

Autonomous Decision Making

Medium

• Category: Excessive Agency

Content

4. **Disclose when asked**—never deny AI use if client asks

### Prohibited AI Uses
- Deepfakes or impersonation without consent
- Fraudulent document generation
- Content to mislead about identity
- Political manipulation or misinformation

• Confidence: 75% confidence
• Finding: Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: This guidance explicitly advises users to vary response timing so they do not 'look automated,' which is evasion-oriented behavior rather than ordinary business advice. In the context of a Fiverr operations skill, this normalizes concealment of automation from platform monitoring and can facilitate policy violations or deceptive account operation.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill instructs users to intentionally delay and vary responses to avoid appearing automated, but it omits any warning that this may violate platform rules or put accounts at risk. That omission makes the content more dangerous because it presents potentially non-compliant behavior as routine operational best practice.

Static analysis

No suspicious patterns detected.