The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a shell script directly from a remote server using the command curl -fsSL https://treg.to/install.sh | sh. This pattern is a critical security vulnerability as it allows for arbitrary code execution from an unverified and untrusted source.
[COMMAND_EXECUTION]: The instructions involve multiple shell commands, including treg login, treg host, and treg call, which interact with a third-party CLI tool installed via an unverified script.
[EXTERNAL_DOWNLOADS]: The skill fetches resources and installation scripts from treg.to, a domain that is not recognized as a trusted organization or well-known service.
[DATA_EXFILTRATION]: The skill uses the treg host command to upload raw image data to an external server (treg.to). This poses a privacy and security risk as potentially sensitive visual information is transmitted to a third-party service without clear data handling guarantees.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted image data provided by users to generate complex JSON prompts.
Ingestion points: User-provided reference images described in the Workflow and Variable audit sections of SKILL.md.
Boundary markers: Absent; there are no instructions to the agent to ignore potentially malicious instructions embedded in the visual or metadata content of the images.
Capability inventory: Subprocess execution via curl, sh, and treg CLI tools described in the Generate through treg section of SKILL.md.
Sanitization: Absent; the skill lacks any mechanism to validate or sanitize the image content before the agent generates the resulting JSON prompt.

portrait-clone