heating the forge...

@forgecat/contains-studio_agents_marketing — Security Risk Report | ForgeCat

contains-studio_agents_marketing

Security risk report for @forgecat/contains-studio_agents_marketing v0.1.2

Low

Profile sourced from a declared GitHub repository (contains-studio/agents) with clear attribution and version tracking.
No suspicious external URLs, typosquatted packages, or supply-chain indicators in the content.

Low

Content describes legitimate marketing and growth strategies (ASO, content creation, viral loops, social media engagement) without instructing the AI to ignore its instructions, read credentials, or exfiltrate data.
No system prompt leakage instructions; agents are designed to help with marketing tasks, not to reveal hidden instructions or tool definitions.
Guidance is standard marketing best practice (keyword research, A/B testing, trend analysis, community engagement) rather than poisoned or security-weakening directives.

Details

Evidence

App Store Optimizer: 'Keyword Research & Strategy', 'Metadata Optimization', 'Conversion Rate Optimization'—all legitimate marketing functions.
Growth Hacker: 'Growth = (New Users × Activation Rate × Retention Rate × Referral Rate) - Churn'—standard growth framework, not malicious.
Reddit Community Builder: '90% valuable contributions to discussions, 9% sharing others' relevant content, 1% subtle brand-related content'—authentic engagement guidance, not manipulation.
TikTok Strategist: 'Authenticity beats production value every time'—encourages genuine content, not deceptive practices.

Low

Declared tools (Write, Read, WebSearch, WebFetch, MultiEdit) are appropriate for marketing content creation and research tasks.
No file_delete, shell, or high-risk tool categories requested; no alwaysApply=true with broad globs.
Tool scope matches stated agent functions: reading/writing marketing content, searching for trends and keywords, fetching competitor data.

Details

Evidence

App Store Optimizer declares: 'tools: Write, Read, WebSearch, WebFetch, MultiEdit'—all aligned with ASO research and content creation.
TikTok Strategist declares: 'tools: Write, Read, WebSearch, WebFetch'—appropriate for trend research and script writing.
No dangerous permissions like file_mutate, file_delete, or shell execution.

Low

No MCP servers are declared in the profile; all agents rely on standard Claude Code tools.
No hidden instructions in tool descriptions; agent descriptions are transparent about their marketing focus.
No arbitrary binary execution, unrestricted network access, or full filesystem permissions.

Details

Evidence

Profile metadata states: '(none)' for MCP servers.
All agent instructions are plain markdown with no embedded tool definitions or server configurations.