ai-creative-strategist

[Skill Scanner] Natural language instruction to download and install from URL detected All findings: [CRITICAL] command_injection: Natural language instruction to download and install from URL detected (CI009) [AITech 9.1.4] [CRITICAL] command_injection: Natural language instruction to download and install from URL detected (CI009) [AITech 9.1.4] This skill's stated purpose (research-led creative strategy with generated visual previews) matches the capabilities it requests (browser scraping, research tools, MCP image-generation). No explicit malicious code is present in this document. Primary risks are: (1) supply-chain — NPX @latest installs and unpinned MCP packages could execute arbitrary code if an upstream package is compromised or a typosquat is used; (2) privacy/data exposure — Playwright and scraping can capture sensitive pages/screenshots and those may be uploaded to remote model providers; (3) credential handling — API tokens stored in local config are sensitive and should be protected. Recommendations: pin NPX package versions, document exact endpoints/providers for the named models (verify 'nano-banana-pro' origin), add explicit privacy and storage handling for screenshots and generated images, and minimize required permissions. Overall I assess this skill as not malicious but carrying moderate supply-chain and data-exposure risk. LLM verification: No evidence of obfuscated malware or explicit malicious code in the SKILL.md. The document is coherent: capabilities align with stated purpose (research + visual previews). Primary risks are supply‑chain (npx dynamic installs from npm), credential handling (storing API tokens in plaintext config files), and privacy/legal exposure from scraping competitor sites and uploading content to third‑party model services. Recommend: verify MCP package publishers, pin versions, avoid storing secrets in wor