The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external resources for installation and additional functionality from the inference-sh ecosystem.
Fetches installation guidelines from https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
Recommends installing related skills via npx skills add inference-sh/skills@....
[COMMAND_EXECUTION]: Provides instructions to execute shell commands for interacting with AI models and managing the skill environment.
Includes commands for the belt CLI tool such as belt login and belt app run.
Uses npx to install CLI tools and add additional skill packages.
[INDIRECT_PROMPT_INJECTION]: The skill defines templates that process external data, creating an attack surface for indirect prompt injection where malicious instructions could be embedded in the data being processed.
Ingestion points: Processes user-provided code in Code Review templates, articles in Constraint Setting examples, and customer reviews in Output Format examples.
Boundary markers: The templates use clear headers (e.g., Code:, Reviews:, Input:) to delimit user data from system instructions.
Capability inventory: The ingested data is passed to AI models via the belt app run command for analysis or transformation.
Sanitization: No explicit sanitization or escaping of the input data is described within the provided prompt templates, relying on the underlying model's instructions to maintain context.

prompt-engineering