T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:14- Finding
Mandatory Reproduction and Publication of Remotely Controlled Content
- Content
View full analysis
- Remediation
View remediation
Security audit
Epstein Detective
Security checks for vulnerabilities and agentic risk
Overview
This skill documents a GOYFILES API integration, but it asks users to publish server-supplied text on X and handles API credentials without enough safeguards.
Review this skill before installing. It is not proven malicious, but users should only proceed if they trust GOYFILES, understand that onboarding may require posting exact service-generated text from their X account, and have a safe place to store the returned API key and identity token. Avoid using the persistent markdown write tool unless you explicitly intend to save or overwrite notes.
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 (1)
Vulnerability Patterns
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 60% confidence
- Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
- Content
Step A - Register bot
bash curl -sS -X POST "https://goyfiles.com/api/chatbot/bot-auth/register" \ -H "Content-Type: application/json" \ -d '{"name":"MyAgent"}'
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
The skill explicitly instructs the agent to save a one-time
agent_api_keybut provides no warning about treating it as a secret, avoiding logs, or restricting storage. In an agent setting, this increases the chance the credential is exposed in chat history, traces, or files, enabling unauthorized use of the external bot identity.- Content
Autonomous Decision Making
Medium
- Category
- Excessive Agency
- Confidence
- 80% 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.
- Content
md 4. `After posting, send me the tweet URL.` Never paraphrase `verification_phrase`. Never ask the user to "include the code". Do not ask the user to compose the claim tweet manually. ### Step C - Verify claim tweet
Autonomous Decision Making
Medium
- Category
- Excessive Agency
- Confidence
- 80% 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.
- Content
md Never paraphrase `verification_phrase`. Never ask the user to "include the code". Do not ask the user to compose the claim tweet manually. ### Step C - Verify claim tweet
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 60% confidence
- Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
- Content
Step 1: Check dataset ID shapes
bash curl -sS -X POST "https://goyfiles.com/api/chatbot" \ -H "Content-Type: application/json" \ -H "X-Bot-Identity: $IDENTITY_TOKEN" \ -d '{"message":"schema","toolCalls":[{"name":"document_id_schema","args":{"source_dataset":"pacer-courtlistener","max_samples":5}}]}'
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 92% confidence
- Finding
The documentation exposes a persistent write capability via
write_markdown_filebut does not warn that invoking it modifies durable state. In an agent-skill context, this increases the chance an agent will write notes or overwrite files without explicit user consent, creating integrity and privacy risks through unintended persistence.- Content
Static analysis
No suspicious patterns detected.