Video Prompting Guide
PassAudited by ClawScan on May 1, 2026.
Overview
This is a coherent AI video prompting guide, but it also points users to install, log into, and run the external inference.sh CLI.
This skill appears safe for learning video prompting. Before using the command examples, verify the inference.sh installer, understand that infsh login uses an external account, and avoid sending sensitive information in generated-video prompts.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent could run inference.sh commands related to the video-generation workflow, sending requests to that service.
The skill allows Bash invocations of the infsh CLI. This is consistent with the inference.sh examples, but users should notice that the agent may run external CLI commands if the tool is installed.
allowed-tools: Bash(infsh *)
Use the skill with clear user confirmation before running generation commands, especially if the CLI is logged into an account.
Running the setup command would execute code downloaded from inference.sh on the user’s machine.
The Quick Start suggests installing an external CLI by piping a remote script to the shell. This is user-directed and disclosed, but it relies on the external installer’s provenance.
curl -fsSL https://cli.inference.sh | sh && infsh login
Prefer the documented manual install and checksum verification path, or inspect the installer before running it.
The CLI may create or use an authenticated inference.sh session when running model requests.
The workflow asks the user to authenticate to inference.sh, even though registry metadata declares no primary credential. This appears purpose-aligned for using the service, but it is account access users should notice.
infsh login
Log in only with the intended account and review the service’s account permissions and logout or token-revocation options.
Video prompt text may leave the local environment and be processed by inference.sh or the selected model provider.
The examples send prompt input to inference.sh-hosted model runs. This external provider flow is disclosed and central to the guide’s examples.
infsh app run google/veo-3-1-fast --input
Avoid including confidential, personal, or proprietary details in prompts unless you are comfortable with the provider’s data handling terms.