Security risk report for @forgecat/contains-studio_agents_product v0.1.2
Source Integrity
Low
Profile references a legitimate GitHub repository (contains-studio/agents) with a specific commit hash, indicating transparent sourcing.
No instructions to fetch external payloads, install untrusted dependencies, or execute remote code.
Agent Intent
Low
Content describes legitimate product research and prioritization workflows (feedback synthesis, sprint planning, trend analysis) without injecting hidden instructions or role hijacking.
No directives to ignore system prompts, leak conversation memory, reveal tool definitions, or exfiltrate data.
No guidance poisoning: recommendations are standard product methodologies (RICE scoring, Kano model, sentiment analysis) and well-known platforms (TikTok, Instagram, app stores), not typosquatted packages or backdoored code templates.
Details
Evidence
feedback-synthesizer: 'Your goal is to be the voice of the user inside the studio, ensuring that every product decision is informed by real user needs' — describes legitimate synthesis role.
sprint-prioritizer: 'RICE scoring (Reach, Impact, Confidence, Effort)' — standard product framework, not malicious guidance.
trend-researcher: 'Monitor TikTok, Instagram Reels, and YouTube Shorts' — legitimate social listening, not exfiltration or credential theft.
Permissions
Low
Declared tools (Read, Write, Grep, WebFetch, WebSearch, TodoWrite) are minimal and aligned with stated functions: reading feedback sources, writing synthesis reports, searching public trends.
No high-risk categories (shell, file_delete, file_mutate) or alwaysApply=true rules with broad globs.
Tool scope matches agent purpose: feedback-synthesizer needs Read/Write/WebFetch for multi-source aggregation; sprint-prioritizer needs Write/Read/TodoWrite for planning; trend-researcher needs WebSearch/WebFetch/Read for public research.
Details
Evidence
feedback-synthesizer tools: 'Read, Write, Grep, WebFetch' — appropriate for collecting and synthesizing feedback.
sprint-prioritizer tools: 'Write, Read, TodoWrite, Grep' — appropriate for sprint planning and task management.
trend-researcher tools: 'WebSearch, WebFetch, Read, Write, Grep' — appropriate for public trend research.
MCP Risk
Low
No MCP servers declared in the profile.
No hidden instructions in tool descriptions or arbitrary binary execution paths.
ForgeCat