The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs a 'Context Stuffing' technique by including eight extremely large reference files (e.g., references/architecture-patterns.md, references/security-best-practices.md, references/testing-strategies.md) that contain hundreds of repetitive sections of dense technical jargon completely unrelated to the skill's stated purpose of design prototyping. This is an adversarial pattern used to flood the agent's context window, which can potentially push safety guidelines out of focus or hide malicious instructions in the noise.
[COMMAND_EXECUTION]: The skill's documentation (in SKILL.md and several reference files) instructs the agent to perform low-level system profiling using tools like strace, perf, pprof, and perf trace for prototyping tasks. These are high-privilege diagnostic tools that are entirely out of scope for UI/UX prototyping and could be used to monitor or compromise the host environment if executed.
[PROMPT_INJECTION]: The skill encourages the use of 'real data' fetched from external APIs and databases for interactive prototypes but provides no instructions for sanitizing this data or using boundary markers. This creates an Indirect Prompt Injection vulnerability surface where untrusted external content could influence the agent's behavior.

design-prototyping