The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to clone a repository from an unverified GitHub user (https://github.com/tt-a1i/archify.git) to obtain its core rendering engine. This source is external and not categorized as a trusted vendor.
[COMMAND_EXECUTION]: The helper scripts render-archify-diagrams.mjs and export-archify-previews.mjs use child_process.spawnSync to execute the Node.js binary, the downloaded archify script, and the system's Google Chrome browser in headless mode.
[REMOTE_CODE_EXECUTION]: By instructing users to download and then automatically executing the 'archify' tool at runtime via its path, the skill facilitates the execution of remote code from an untrusted source.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection if diagram JSON files are sourced from untrusted environments, as these files are rendered into HTML and opened in a browser context.
Ingestion points: render-archify-diagrams.mjs collects and processes *.json diagram files from directories specified by the user or configuration.
Boundary markers: None identified; the skill does not use delimiters or warnings to prevent the agent from obeying instructions embedded within diagram labels or metadata.
Capability inventory: The skill executes shell commands (spawnSync), writes files to the filesystem, and opens a headless browser instance with access to the file:// protocol.
Sanitization: There is no evidence of sanitization for the diagram JSON content before it is interpolated into HTML templates and rendered by the browser.

soia-dev-archify-diagrams