The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external programs through the shell and Python's subprocess module. Specifically, audit_partners.py runs curl to verify the accessibility of partner URLs. The SKILL.md file also provides instructions for using sed to sanitize SVG files and rsvg-convert to rasterize images. These commands are essential for the skill's auditing and image processing features.
[EXTERNAL_DOWNLOADS]: The skill interacts with external sources by probing partner homepages via curl to check for broken links. It also depends on the lovstudio-find-logo skill for brand discovery and requires the installation of the Pillow library and librsvg package. These dependencies are standard for the described image manipulation and auditing tasks.
[PROMPT_INJECTION]: The skill is designed to ingest and process user-provided data, which represents an indirect prompt injection surface.
Ingestion points: Brand names, slugs, and homepage URLs are collected from user input (via AskUserQuestion) and then incorporated into the website's source code and configuration files.
Boundary markers: There are no explicit boundary markers or instructions to isolate user-provided strings during script execution.
Capability inventory: The skill possesses the ability to modify local repository files (add_partner.py) and perform network requests (audit_partners.py).
Sanitization: While the scripts utilize argparse for handling inputs, the agent must ensure that user-supplied strings, particularly slugs, do not contain path traversal sequences or shell metacharacters before executing the associated scripts.

lovstudio-maintain-partners