brand-guidelines — agentic threat model

Not certain from the listing — relies on an underlying foundation model (presumably Anthropic Claude) to interpret styling requests and apply typography logic. Vulnerable to prompt injection that could bypass brand guidelines to output malicious HTML/CSS payloads.

The agent uses static brand assets (hex codes, font names) and does not appear to maintain a dynamic vector database or ingest external RAG data, limiting data poisoning vectors.

Orchestrates styling logic over documents, slides, and HTML. Risks include insecure tool integration if the post-processing engine parses untrusted user documents or executes arbitrary styling scripts.

Not certain from the listing — the deployment infrastructure hosting this skill is unspecified. If run in an unsandboxed environment, the document post-processing logic could be targeted for local file access or remote code execution.

Not certain from the listing — there is no mention of built-in evaluation, logging, or guardrails to verify that the generated HTML/styling outputs do not contain malicious injections or cross-site scripting (XSS) vectors.

Not certain from the listing — compliance controls, access policies, and audit logging are not detailed. The open-source nature suggests security responsibility is deferred to the deploying party.

Designed as a 'skill' which implies it is called by other agents or frameworks within an ecosystem. Vulnerable to downstream exploitation if a compromised orchestrator agent feeds it malicious inputs to style.