The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructional constraints like 'CRITICAL: If no design system exists, ask before creating one' and 'NEVER: Extract one-off...'. These are task-specific guardrails designed to maintain code quality and follow user preferences, rather than attempts to bypass system safety filters. Additionally, while the skill ingests untrusted codebase data (Ingestion points: grep discovery in SKILL.md), this surface for indirect prompt injection is a common characteristic of development-oriented skills and is considered safe within this context.
[DATA_EXFILTRATION]: The skill performs local file analysis to identify design patterns. It does not target sensitive file paths (e.g., credentials, SSH keys) or involve network operations that could facilitate data exfiltration.
[COMMAND_EXECUTION]: The skill utilizes standard search utilities such as grep to analyze the project structure. These commands are used for their intended purpose in a development workflow and do not involve privilege escalation or unauthorized system modifications.

extract