The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the installation and update of the infsh CLI through a shell pipe pattern (curl -fsSL https://cli.inference.sh | sh). This method executes remote code without prior inspection, creating a significant risk of arbitrary code execution if the remote server or the network connection is compromised.
[EXTERNAL_DOWNLOADS]: The skill downloads various resources, including platform-specific binaries, checksum files, and metadata manifests, from dist.inference.sh and cli.inference.sh during the setup and update processes.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute infsh commands, which involves managing system processes, local authentication, and network-based AI task execution.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by accepting user-supplied prompts and JSON data for processing by various AI applications.
Ingestion points: User prompts and input files are ingested via the infsh app run --input command as described in SKILL.md and references/running-apps.md.
Boundary markers: No specific boundary markers or "ignore embedded instructions" warnings are implemented to protect the tool input.
Capability inventory: The infsh tool can perform network operations, file writing (via --save), and access remote AI services (e.g., Twitter/X, search engines).
Sanitization: There is no evidence of input validation or sanitization before external content is passed to the underlying CLI tool.

agent-tools