The Agent Skills Directory

[SAFE]: The documented functions are standard Node.js utilities for backend infrastructure management.
[DATA_EXFILTRATION]: The skill provides methods to access system-level information, including network interfaces via getNetworkInterfaces (generated/system/network.d.ts) and environment variables via loadEnv (generated/config/env.d.ts). These are standard features for its intended use case.
[COMMAND_EXECUTION]: Documentation for getDiskInfo (generated/system/disk.d.ts) specifies that it executes PowerShell commands on Windows to retrieve disk metrics.
[PROMPT_INJECTION]: The skill documents interfaces for ingesting untrusted data through configuration and environment files, creating an indirect prompt injection surface. 1. Ingestion points: loadConfig (generated/config/config.d.ts) and loadEnv (generated/config/env.d.ts). 2. Boundary markers: No delimiters or instruction-ignore warnings are identified in the API definitions. 3. Capability inventory: File system modification via writeFile (generated/fs/write-file.d.ts) and removePath (generated/fs/remove.d.ts), and autonomous task execution via the Scheduler class (generated/scheduler/scheduler.d.ts). 4. Sanitization: No input validation or content filtering is implemented in the provided library definitions.

cat-kit-be