heating the forge...

@forgecat/msitarzewski_agency-agents_agency-design — Security Risk Report | ForgeCat

msitarzewski_agency-agents_agency-design

Security risk report for @forgecat/msitarzewski_agency-agents_agency-design v0.0.11

Low

Profile sourced from public GitHub repository (msitarzewski/agency-agents) with MIT license.
Original commit hash and version tracking provided for supply-chain transparency.
No evidence of typosquatted dependencies, backdoored code templates, or attacker-controlled package recommendations.

Low

Content describes design agent personalities, workflows, and deliverable templates—no instructions to manipulate the AI, ignore system prompts, or exfiltrate data.
Markdown files document legitimate design methodologies (brand strategy, UI systems, UX research, visual storytelling) with no hidden directives.
No guidance poisoning detected: recommendations for tools, libraries, and practices are standard industry patterns (CSS design tokens, accessibility standards, typography scales, responsive design frameworks).

Details

Findings

Evidence

Brand Guardian: 'Create comprehensive brand foundations' and 'Guard brand consistency'—legitimate design strategy, not prompt injection.
Image Prompt Engineer: Detailed photography terminology and prompt structure—educational content for AI image generation, not instruction to manipulate the agent.
Inclusive Visuals Specialist: Technical constraints to counter AI bias (e.g., 'No Clone Faces', 'No Gibberish Text')—describes bias-mitigation methodology, not instruction to hide or deny the agent's own instructions.
UI Designer & UX Architect: CSS design systems, component libraries, responsive frameworks—standard web design practices, not security-weakening guidance.
Visual Storyteller: Narrative structure, storyboarding, multimedia frameworks—creative methodology, not malicious knowledge injection.

Low

No tool definitions or MCP servers declared in profile.
No shell, file_write, file_delete, or high-risk tool categories requested.
Profile is pure markdown documentation of agent personalities and design methodologies—no executable permissions or authority claims.

Details

Evidence

Profile metadata: 'Dependencies: None'.
No tool configurations, hooks, or alwaysApply rules present.
Content is instructional/reference material for human designers and AI agents, not a tool manifest.

Low

No MCP servers defined in profile.
No binary execution, network access, or filesystem permissions requested.
Profile contains only markdown documentation and design templates—no server configuration or external integration.

Details

Evidence

Profile section 'MCP servers (untrusted data): (none)'.
No tool descriptions containing hidden instructions or arbitrary code execution.
No environment credentials or external host dependencies declared.