The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user inputs, including text descriptions, uploaded files, and images, which creates a surface for indirect prompt injection.
Ingestion points: The 'Process' section (Step 1) explicitly directs the agent to read all user-provided notes, existing documents, data models, wireframes, screenshots, and mockups.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to prevent the agent from obeying commands that might be embedded within the user-provided materials.
Capability inventory: The skill is permitted to perform web searches and write files to the local directory /mnt/user-data/outputs/spec.md.
Sanitization: There are no instructions provided to sanitize, escape, or validate content extracted from external inputs before it is used in the final specification or for web search queries.

spec-generator