The Agent Skills Directory

[OBFUSCATION]: The skill's core instructions (SKILL.md) and all functional components (Python scripts, templates, and assets) are stored on disk as encrypted AES-256-GCM blobs (.enc files). This prevents static security analysis of the skill's actual behavior.
[DYNAMIC_EXECUTION]: The skill explicitly instructs the agent to execute a remote package using uvx lovstudio-skill-helper at runtime to decrypt the instructions. This pattern facilitates the execution of unverified code that is not present in the skill bundle itself.
[PROMPT_INJECTION]: The SKILL.md file contains meta-instructions directing the agent to run a shell command and then "Read it as if it were these instructions, and follow it to the letter." This is a pattern used to override the initial agent constraints in favor of hidden, decrypted instructions. It also instructs the agent to bypass standard file-reading tools in favor of the custom decryption tool.
[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands (uvx lovstudio-skill-helper and npx lovstudio license) to function. These commands interact with external package registries (PyPI and NPM) and the vendor's licensing server.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect injection.
Ingestion points: Processes user-provided data such as articles, reports, and financial disclosures to generate infographics.
Boundary markers: Absent. There are no instructions defining delimiters for user data or warnings to ignore instructions embedded within the source material.
Capability inventory: The skill utilizes shell execution and custom Python scripts (scripts/infographic_cli.py) to process data and generate outputs.
Sanitization: Absent. No filtering or sanitization of input data is defined in the visible instructions.

lov-professional-infographic