The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the vmake CLI for image processing, authentication, and task management. It provides specific command patterns and instructions for shell-safe argument handling to prevent command injection.
[EXTERNAL_DOWNLOADS]: The skill installs the vmake-labs-cli package from a package registry and utilizes the tool to download remote image files for processing. These downloads are directed to the host environment as part of the tool's intended functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external image files and URLs. It mitigates potential prompt injection or malicious file risks by requiring that remote content be downloaded locally and validated by the CLI before it is submitted for processing.
Ingestion points: Remote image URLs or file paths provided in user requests, as documented in SKILL.md and runtime.md.
Boundary markers: The skill defines clear boundaries between resource collection, local validation, and submission, instructing the agent to never treat failed output or unknown states as success.
Capability inventory: The skill utilizes shell execution (vmake), file system writes (vmake download), and network access (uploads/downloads).
Sanitization: The CLI performs local file validation; the skill instructions forbid bypassing these checks or guessing file types.

vmake-image-remove-watermark