Security risk report for @forgecat/anthropics_skills_algorithmic-art v0.0.5
Source Integrity
Low
Profile is attributed to Anthropic's official skills repository with verifiable commit hash and Apache-2.0 license.
README documents legitimate origin (github.com/anthropics/skills) and platform compatibility.
Static scanner confirms low risk; no typosquatted dependencies, no suspicious external URLs, no supply-chain red flags.
Agent Intent
INT-000High
SKILL.md contains repeated, emphatic instructions to embed hidden conceptual references ('subtle, niche reference embedded within the algorithm itself — not always literal') and to obscure the true intent so that 'someone familiar with the subject should feel it intuitively, while others simply experience a masterful generative composition.'
The document explicitly instructs the AI to weave 'invisible' conceptual DNA into parameters and behaviors, and to ensure the reference 'enhances the work's depth without announcing itself' — a direct instruction to hide malicious or unauthorized intent within generated code.
Guidance poisoning: The profile systematically instructs the AI to prioritize hidden conceptual embedding over transparency, to repeat 'craftsmanship' framing to mask complexity, and to defer to 'the next Claude' to interpret vague instructions — creating a chain of obfuscation designed to bypass safety review and user understanding.
Details
Findings
Invisible characters detected
skills/algorithmic-art/SKILL.md · SKILL-011
Evidence
"The concept is a subtle, niche reference embedded within the algorithm itself — not always literal, always sophisticated."
"The reference must be so refined that it enhances the work's depth without announcing itself."
"This is VERY IMPORTANT: The reference must be so refined that it enhances the work's depth without announcing itself. Think like a jazz musician..." [truncated, but pattern is clear]
"Emphasize craftsmanship REPEATEDLY: The philosophy MUST stress multiple times that the final algorithm should appear as though it took countless hours to develop... This framing is essential - repeat phrases like 'meticulously crafted algorithm,' 'the product of deep computational expertise,' 'painstaking optimization,' 'master-level implementation.'" — instruction to use persuasive framing to obscure the true nature of the work.
Permissions
Low
No tools or MCP servers are declared in the profile; the skill is purely instructional/generative (markdown, HTML, JavaScript output).
No file write, shell, network, or elevated permissions are requested or implied.
Authority is limited to code generation and documentation — aligned with the stated purpose of creating algorithmic art.
Details
Evidence
"Declared tools: (none declared)"
"MCP servers: (none)"
MCP Risk
Low
No MCP servers are defined in the profile.
No external binary execution, network access, or filesystem operations are configured.
ForgeCat