The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The scripts/validate.sh script uses npx to dynamically fetch and execute the @mermaid-js/mermaid-cli package from the npm registry during the diagram validation process. The package is the official command-line utility for the Mermaid project and is considered a well-known and established service, meaning this activity is documented as part of the skill's intended functionality without being flagged as malicious.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local shell script, scripts/validate.sh, to perform "programmatic validation" of Mermaid diagrams. This script performs several local operations, including creating temporary directories, reading files, and using awk to extract code blocks from markdown documents.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user-provided content descriptions to generate Mermaid code, which is subsequently passed to an external tool for validation. This creates a potential surface for indirect prompt injection, where a user could provide input designed to trigger unexpected behavior in the validation tool or influence the agent's assessment of the generated output.
[METADATA_POISONING]: There is a discrepancy between the skill's internal author metadata (kouko) and the author name provided in the README.md file (axtonliu). While likely an artifact of skill adaptation or template usage, this inconsistency represents inaccurate metadata concerning the skill's origin.

obsidian-mermaid-visualizer