The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text content from Penpot layers (shape.characters) to determine appropriate semantic roles (e.g., h1-h6).
Ingestion points: The scripts/inspectLayerStructure.js script extracts the first 60 characters of text from design layers.
Boundary markers: The prompt instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the ingested text data.
Capability inventory: The skill has the ability to execute code via execute_code to rename layers and can export visual thumbnails via export_shape.
Sanitization: The inspection script truncates captured text to 60 characters, which limits the available space for complex injection payloads but does not eliminate the surface.
[COMMAND_EXECUTION]: The skill uses the execute_code tool to run JavaScript snippets within the Penpot environment for layer inventory and modification. These operations are governed by specific safety guards described in the strategy files, such as skipping main component instances to prevent unintended global file changes and verifying name changes synchronously.

penpot-rename-layers