The Agent Skills Directory

[DYNAMIC_EXECUTION]: The shell scripts generate-app-icons.sh and search-icons.sh utilize heredocs to dynamically generate and execute Python code. In generate-app-icons.sh, the variable $ICON_NAME is interpolated directly into the Python source code without prior validation or sanitization. This pattern could theoretically allow for arbitrary code execution within the Python process if a malicious icon ID were processed.
[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands to manage Android project assets. This includes using find ... -delete to remove legacy raster icons and executing gradlew for building the project. These operations affect the local filesystem and project state.
[EXTERNAL_DOWNLOADS]: The skill fetches icon data, metadata, and SVG files from the Iconify API (api.iconify.design). These downloads are essential for the skill's primary function and are documented neutrally as they target a well-known service for icon assets.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from an external API. While basic numeric validation is performed on icon dimensions, the SVG path data and icon identifiers are processed via regex and local scripts without comprehensive sanitization, creating a potential surface for injection via malicious API responses.
[PRIVILEGE_ESCALATION]: The SKILL.md file includes instructions for the user to install system dependencies using sudo apt install. Although these commands are documentation for the user rather than automated scripts, they involve the request for elevated permissions on the host system.

android-app-icon