The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the supaguard CLI for authentication and check management, and performs a search for installed companion skills on the filesystem.
[REMOTE_CODE_EXECUTION]: The skill performs dynamic generation of Playwright scripts which are then executed on a remote cloud runner through the supaguard CLI.
[EXTERNAL_DOWNLOADS]: The skill directs the user to install the supaguard CLI package from the public npm registry.
[DATA_EXFILTRATION]: Application logic and route structures are extracted from the local source code and transmitted to the supaguard platform during the check creation and testing process.
[PROMPT_INJECTION]: The skill analyzes untrusted application code to generate monitoring logic, which presents a surface for indirect prompt injection.
Ingestion points: Processes codebase directories including app, pages, and src to identify routes and UI selectors.
Boundary markers: Lacks explicit delimiters or instructions to the agent to distinguish between its own logic and the analyzed code fragments.
Capability inventory: The skill can write files to the /tmp directory and execute system commands via the CLI.
Sanitization: Implements a sandbox policy that forbids sensitive Node.js modules (such as fs, child_process, and net) and unsafe functions like eval in the generated scripts.

supaguard