Security audit

Collabs & Cross-Promotion

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malicious, but it needs review because it guides coordinated public social posting while its giveaway and native-collab scope is inconsistent.

Review this skill before installing. It is suitable as a planning and drafting aid, but require explicit human approval from each account owner before anything is scheduled or published. Treat giveaways, duet/stitch, IG/YouTube native collaborations, and legal disclosure decisions as outside WoopSocial automation unless separately verified against platform rules and applicable law.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The eval at line 42 expects the agent to recommend duet/stitch and joint giveaway formats even though the skill description explicitly says this skill does not handle duet/stitch or giveaway mechanics. This contradiction can push the agent to produce out-of-scope guidance, weakening safety boundaries and causing downstream misuse or user deception about platform capabilities.

Description-Behavior Mismatch

Medium

Confidence: 96% confidence
Finding: The reference file lists formats the manifest explicitly says the skill does not support, including duet/stitch, joint giveaway, and native YouTube/IG collaboration variants. This creates specification drift: an agent using the reference may recommend or operationalize prohibited workflows, leading users into unsupported actions, policy noncompliance, and unreliable execution.

Description-Behavior Mismatch

High

Confidence: 99% confidence
Finding: The worked example gives concrete instructions for running a joint giveaway even though the manifest says the skill does not run giveaways. Because worked examples are high-authority guidance for downstream agents, this is more dangerous than a generic mention: it can directly induce unsupported promotional behavior and incorrect compliance decisions, including the oversimplified claim that no payment means no #ad.

Missing User Warnings

Medium

Confidence: 85% confidence
Finding: The skill explicitly states that WoopSocial publishes each partner's own posts on a coordinated schedule, but it does not mention any operator confirmation, access control, consent capture, or rollback safeguards for account-impacting actions. Because publishing to social accounts can directly affect brand reputation, timing-sensitive campaigns, and account integrity, omission of these safeguards creates a real risk of unauthorized or mistaken posting.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The coordinated publishing guidance instructs simultaneous live posting across accounts but does not warn about scheduling errors, accidental publication, reputational harm, or the need for final approval prior to posting. In a collab workflow, a bad schedule or incorrect post can impact both parties at once, amplifying harm beyond a single-account mistake.

Static analysis

No suspicious patterns detected.