The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains script templates that ingest external parameters such as length and character sets into shell commands, creating a surface for indirect prompt injection.\n
Ingestion points: File SKILL.md contains bash script examples using unvalidated variables $LENGTH, $CHARS, and $PREFIX.\n
Boundary markers: No delimiters or 'ignore embedded instructions' warnings are used in the prompt templates.\n
Capability inventory: Uses openssl, tr, sed, shuf, python3, and node to perform operations on the local system.\n
Sanitization: No input validation, escaping, or filtering is implemented in the provided shell script templates to prevent command injection via parameters.\n- [COMMAND_EXECUTION]: The skill provides commands for local execution of standard system utilities. While these are intended for legitimate cryptographic generation, they represent the capability to execute code on the host environment.\n- [SAFE]: No network access, external downloads, or attempts to access sensitive files were detected. All random generation uses cryptographically secure sources like /dev/urandom and openssl rand.

password-gen