The Agent Skills Directory

[SAFE]: The skill provides a clear, collaborative framework for architectural discussions and design planning without requesting or executing dangerous capabilities.
[COMMAND_EXECUTION]: The skill contains explicit restrictions ('Hard Gates') preventing the agent from writing code or scaffolding projects, which effectively mitigates risks associated with unauthorized file system modifications.
[DATA_EXPOSURE]: Although the skill workflow involves reading repository context (files, docs, configs), this is a standard requirement for its design-assistance purpose and no patterns of data exfiltration or credential harvesting were found.
[PROMPT_INJECTION]: The instructions focus on structured dialogue and design synthesis; no patterns attempting to override safety filters, extract system prompts, or bypass agent constraints were identified.

code-brainstorming