The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied keywords and reads local project files (.styleseed/project.json, .styleseed/artifacts/index.json) to determine how to modify source code. While this creates a surface for potential instruction injection if project files are malicious, the skill uses these inputs within a restricted logic for UI component modification.
Ingestion points: Processes user input (vibe/context) and content from local project configuration files and target JSX files.
Boundary markers: None explicitly defined to separate untrusted data from instructions.
Capability inventory: Possesses file read/write/edit and bash capabilities via allowed-tools.
Sanitization: No explicit sanitization or validation of the input data is described, though logic is restricted to a fixed lookup table for motion seeds.
[COMMAND_EXECUTION]: The skill has access to the Bash tool. The instructions do not contain specific shell commands but the tool is listed in the frontmatter to facilitate general environment resolution and file operations required for code editing.

ss-motion