The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to interact with the environment. Specifically, it uses mcporter (an MCP bridge tool) to call the Higgsfield image generation service and magick (ImageMagick) to assemble contact sheets from generated image files. These commands are critical to the skill's primary function.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data including book chapters, transcripts, and user-supplied ideas. This data is transformed into prompts which are subsequently used as arguments in shell-based tool calls, creating an injection surface.
Ingestion points: Raw user input in SKILL.md and document expansion in references/method.md.
Boundary markers: The skill uses "No other text" constraints in the generated prompts to guide the model, though these do not act as strict technical sanitization for the execution environment.
Capability inventory: The skill requires access to shell execution (mcporter, magick) and network access via curl to download generated assets.
Sanitization: While the instructions suggest quoting values containing spaces in shell calls, there is no formal logic to escape or filter malicious control characters within the user-derived content.
[EXTERNAL_DOWNLOADS]: The skill uses curl to download image files from remote URLs. These downloads target the Higgsfield service—a well-known generative AI provider—as an integral part of the image generation workflow. The references are documented neutrally as expected for a well-known service interaction.

idea-to-visual