The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted markdown data which creates a vulnerability surface where malicious code within Mermaid diagrams or section headers could attempt to influence the processing logic.
Ingestion points: The main bash script extract-and-generate.sh accepts a user-provided markdown file path as its primary input.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the processed markdown content.
Capability inventory: The skill invokes several subprocesses, including python3, mmdc (which launches a headless browser), and identify from the ImageMagick suite.
Sanitization: The Python script extract_diagrams.py uses regular expressions to sanitize headers before using them as filenames, but the core diagram content is passed directly to the mmdc renderer without validation or sanitization.
[COMMAND_EXECUTION]: The extract-and-generate.sh script executes multiple system commands to manage directories, process data, and validate generated images, including mkdir, python3, mmdc, stat, and identify.
[EXTERNAL_DOWNLOADS]: The documentation in references/setup_and_troubleshooting.md provides commands to fetch the Google Chrome signing key and installation packages from Google's official servers.
[PRIVILEGE_ESCALATION]: Setup instructions in the troubleshooting guide include the use of sudo for repository configuration and the installation of system-level packages required for the Chrome browser and its dependencies.

mermaid-tools