Security audit
Security checks for vulnerabilities and agentic risk
This skill is not malicious, but it should be reviewed because it can run unpinned SocQ tooling with an API key and exposes broad public-data collection beyond its social/SEO framing.
Install only if you intend to let an agent send queries, URLs, usernames, and other public research targets to SocQ using your SOCQ_API_KEY. Scope requests tightly, confirm paid or bulk runs, avoid unnecessary people-focused collection, and prefer a pinned or otherwise vetted SocQ CLI/MCP version before using npx-based fallbacks.
Referenced artifact was not completely inspected
- Search the live Capability Registry first. Read [catalog.md](references/catalog.md) when tool discovery is unavailable.
The skill metadata permits execution via npx without pinning an exact package version, which creates a supply-chain risk: future or compromised package releases could be fetched and run at execution time. Because this skill is designed to handle an API key and invoke external tooling, an attacker controlling the package or dependency path could execute arbitrary code and exfiltrate credentials or research data.
The fallback instruction to use npx @socq/cli has the same unpinned runtime execution risk as the MCP invocation. If the package ecosystem entry is compromised, the agent may run attacker-controlled code in a context that includes network access and the SOCQ_API_KEY secret.
The fallback instruction to use npx @socq/cli has the same unpinned runtime execution risk as the MCP invocation. If the package ecosystem entry is compromised, the agent may run attacker-controlled code in a context that includes network access and the SOCQ_API_KEY secret.
The manifest says to use the skill for 'this public social or SEO research request,' which is a broad natural-language description rather than a narrowly scoped trigger. It does not define clear boundaries or negative examples, so the skill could be invoked for a wide range of common research-related requests beyond the intended scope.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
value: "socq"
description: "Hosted SocQ social and SEO data tools"
transport: "streamable_http"
url: "https://api.socq.ai/mcp"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
value: "socq"
description: "Hosted SocQ social and SEO data tools"
transport: "streamable_http"
url: "https://api.socq.ai/mcp"
The catalog advertises large-scale public data collection across many platforms without any visible warning about privacy expectations, sensitive data handling, or acceptable-use boundaries. In a skill designed for agent consumption, this omission makes misuse more likely because operators are not prompted to consider collection limits, retention, or downstream handling of scraped personal data.
The catalog materially exceeds the declared 'social research' scope by exposing commerce, marketplace, maps, and job-intelligence endpoints. This scope expansion increases the chance that an agent or user can repurpose the skill for unrelated data collection or surveillance workflows that were not clearly disclosed or constrained.
General e-commerce and marketplace scraping endpoints are not justified by the stated social/SEO research purpose, creating unnecessary capability exposure. Even if the data is public, exposing unrelated collection surfaces enables broader profiling and commercial intelligence gathering than users would reasonably expect from this skill.
The file explicitly enables collection of public Facebook comments, profiles, photos, reviews, events, and posts, but provides no warning about privacy sensitivity, lawful use, retention, or handling constraints. Even though the data is described as public, aggregating and normalizing social-platform content can materially increase privacy risk, enable profiling, and lead agents to collect personal data without appropriate safeguards.
This file documents endpoints that enable collection of public Instagram relationship data, profile metadata, posts, comments, reels, and transcripts, including follower/following lists that can be privacy-sensitive when aggregated at scale. Even though the data is described as public, the skill provides no warning, consent boundary, or misuse caution, which increases the risk of surveillance, profiling, or targeting by downstream agents or users.
This file documents collection of public LinkedIn profiles, posts, comments, people-search results, and company/job data without any privacy, consent, or acceptable-use guidance. Even when data is publicly accessible, aggregating and operationalizing it at scale can enable profiling, surveillance, and downstream misuse, so omission of safeguards is a real security/privacy weakness in a research skill.
The file repeatedly states that the default language is English, which imposes a specific language/locale preference in the skill guidance. Because the documentation does not frame this as user-selectable at invocation time, it can steer use toward English without explicit user opt-in.
The file documents collection of public TikTok data including follower graphs, comments, profiles, live status, and transcripts, but does not warn users that invoking these endpoints retrieves external third-party data and may raise privacy, compliance, or data-handling obligations. In a research-oriented skill, that omission can lead downstream agents to collect sensitive social data by default without informed user consent, purpose limitation, or minimization.
This file documents endpoints that enable collection of public social-graph and content data such as followers, following, mentions, search results, thread context, and user posts, but it does not include any privacy, acceptable-use, or data-handling cautions. Even though the data is described as public, aggregating and normalizing it at scale can facilitate profiling, surveillance, or targeting, so omission of warnings increases misuse risk by downstream agents and users.
The file documents YouTube endpoints and example inputs but does not warn that user-provided queries, URLs, hashtags, comment IDs, and related identifiers will be transmitted to the external SocQ service. This creates a privacy and data-handling risk because operators may unknowingly send sensitive investigation terms, targets, or internal research context to a third party.
Google Maps place and review collection is adjacent to research use cases, but it is still outside the declared social-platform and SEO scope in the manifest. This mismatch is lower severity than commerce scraping, yet it still broadens the skill into location/business intelligence without clear disclosure.
LinkedIn jobs and job-search endpoints introduce employment-intelligence capabilities that are distinct from social-content and SEO research. That additional surface could be used for organizational profiling or recruiting surveillance beyond the user's expected purpose for this skill.
This file documents endpoints that collect public Rednote profiles, notes, comments, and recommendations from external services, but it does not warn users that using the skill transmits identifiers, URLs, and queries to third-party endpoints and may retrieve personal or profile-related data. While the data is described as public, the omission can still mislead users about privacy, consent, and compliance expectations when performing social research at scale.
The instruction says to read this file whenever a request 'targets Snapchat,' but it does not define what counts as targeting Snapchat or provide exclusion examples. This ambiguity could cause unintended invocation for loosely related requests mentioning Snapchat without needing these collection endpoints.
SQP-3 applies to all file types and covers language or locale policy violations. The example for the transcript tool specifies "language": "en", which can imply a forced language setting without user opt-in or explanation of why English is required.
SQP-3 covers language or locale policy violations across all file types. The example for youtube/transcripts specifies "language": "en", which can bias the skill toward English without indicating that users may choose another language or that English is merely an example.