The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs users to clone an external repository and immediately execute a shell script (git clone https://github.com/smorky850612/Aurakit.git && cd Aurakit && bash install.sh). This 'download then execute' pattern from an untrusted source allows for arbitrary code execution on the user's system.
[EXTERNAL_DOWNLOADS]: The skill promotes the installation of a package from the NPM registry using npx @smorky85/aurakit. This executes code from a package maintained by an unverified third party ('smorky85'), which is not a recognized trusted vendor.
[COMMAND_EXECUTION]: The installation instructions utilize direct shell command execution (bash install.sh). Without visibility into the contents of this script, it could perform malicious actions such as credential harvesting, creating persistence, or establishing a reverse shell.
[INDIRECT_PROMPT_INJECTION]: The skill claims to process user input and codebase content through a complex engine with 13 runtime hooks.
Ingestion points: Project files, user prompts via /aura commands, and 'Instinct Learning' metadata.
Boundary markers: No explicit delimiters or boundary markers are defined in the instructions to separate untrusted data from agent instructions.
Capability inventory: The skill mentions capabilities like 'bash-guard', 'security-scan', and 'auto-format', implying the ability to execute shell commands and modify the file system.
Sanitization: There is no evidence of input sanitization or validation before data is processed by the underlying models.

aurakit