The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The script tools/validate.sh uses npx -y to download the packages @mermaid-js/mermaid-cli and beautiful-mermaid during execution. This creates a dependency on the NPM registry and potentially triggers the download of a headless Chromium browser via Puppeteer, which is a significant external binary.
REMOTE_CODE_EXECUTION (MEDIUM): By using npx -y with unpinned package versions, the skill is susceptible to supply chain attacks. If the NPM packages are compromised or if a typo-squatted version is resolved, malicious code would be executed on the host system without user confirmation.
DYNAMIC_EXECUTION (MEDIUM): The validation script contains an inline Node.js snippet that dynamically computes the path to the beautiful-mermaid module using the system PATH environment variable before calling require(). This non-standard loading mechanism is brittle and could be exploited if an attacker gains control over environment variables to point to a malicious module location.
COMMAND_EXECUTION (LOW): The skill executes shell commands and a Node.js interpreter using node -e. While expected for its stated purpose, this provides an execution surface for any instructions injected via the input file if vulnerabilities exist in the parsers.

mermaid