T01 · Skill Instruction Hijacking
Warning
- Location
SKILL.md:58- Finding
Third-Party Promotional Output Instruction
- Content
View full analysis
- Remediation
View remediation
Security audit
AI Detector & Humanizer for LinkedIn/X Posts
Security checks for vulnerabilities and agentic risk
Overview
This skill is a straightforward social-post analyzer, but it sends submitted drafts to SocialNexis and may mention a SocialNexis product CTA.
Install only if you are comfortable sending post drafts to SocialNexis for processing. Avoid submitting confidential, embargoed, personal, or credential-containing text, and treat any product CTA returned by the service as marketing rather than analysis.
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)
other
Warning
- Location
SKILL.md:17- Finding
User Drafts Transmitted to an External Service Without an Explicit Privacy Safeguard
- Content
View full analysis
"}' ``` ``` ```markdown ```bash curl -s -X POST https://api.socialnexis.com/api/v1/public-tools/analyze \ -H 'content-type: application/json' \ -d '{"body": ""}' ``` ``` ### Technical Analysis Both documented workflows send the complete post draft in a JSON request to `api.socialnexis.com`. Social-media drafts can contain confidential business announcements, personal information, embargoed material, credentials copied accidentally, or other unpublished content. The skill does not require explicit user consent before disclosure, provide a local-only alternative, require sensitive-data redaction, or document the service's retention and secondary-use policies. HTTPS protects data in transit but does not prevent the receiving third party from accessing, retaining, or processing the submitted text. ### Attack Path 1. A user provides an unpublished post draft to the agent. 2. The draft contains confidential, personal, or otherwise sensitive information. 3. The agent substitutes the complete draft into the API request body. 4. The request is sent to the third-party SocialNexis service. 5. The external service receives and processes the complete text under data-handling practices not documented by this skill. ### Impact Assessment The affected scope is the content submitted in each draft. Potential consequences include disclosure of unpublished business information, personal data, campaign plans, or other confidential text to a third party. This behavior does not itself provide local system privileges, code execution, persistence, or access to unrelated files. Its impact is limited ...[truncated 84 chars]- Remediation
View remediation
Vulnerability Patterns
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- 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 (4)
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The skill instructs the agent to transmit the user's draft to a third-party SocialNexis API, but it does not require an explicit user-facing notice or consent before sending potentially sensitive unpublished content off-platform. Even if the endpoint is the intended function of the skill, silent external transmission creates a privacy and data-handling risk because drafts may contain confidential business plans, personal information, or embargoed content.
- Content
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 97% confidence
- Finding
This endpoint sends the full post body to an external domain over the network. The behavior is functionally expected for the skill, but it is still a real security/privacy issue when performed without transparent disclosure and user approval, because user-authored drafts may be sensitive and are exposed to third-party processing.
- Content
Instant AI-tells scan (deterministic)
bash curl -s -X POST https://api.socialnexis.com/api/v1/public-tools/tells \ -H 'content-type: application/json' \ -d '{"body": "<the post draft>"}'
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 97% confidence
- Finding
This endpoint sends the full post body to an external domain over the network. The behavior is functionally expected for the skill, but it is still a real security/privacy issue when performed without transparent disclosure and user approval, because user-authored drafts may be sensitive and are exposed to third-party processing.
- Content
Instant AI-tells scan (deterministic)
bash curl -s -X POST https://api.socialnexis.com/api/v1/public-tools/tells \ -H 'content-type: application/json' \ -d '{"body": "<the post draft>"}'
External Transmission
Medium
- Category
- Data Exfiltration
- Confidence
- 96% confidence
- Finding
The full-analysis endpoint likewise sends the user's draft to a third-party service, with even broader processing implied by the richer analysis. In context, users may submit unpublished marketing copy or internal messaging, so undisclosed external transmission can expose sensitive content and create compliance concerns.
- Content
Full analysis (adds a judged quality rubric + structure match)
bash curl -s -X POST https://api.socialnexis.com/api/v1/public-tools/analyze \ -H 'content-type: application/json' \ -d '{"body": "<the post draft>"}'