The Agent Skills Directory

[SAFE]: The skill operates as a high-level router for interview-driven planning tasks, delegating logic to specific sub-skills like grilling and domain-modeling.- [COMMAND_EXECUTION]: Explicit safety constraints are defined for the sdd mode, which strictly limits the agent to read-only codebase exploration and prohibits code edits, builds, installs, or any state-changing shell commands.- [DATA_EXPOSURE]: File system activity is restricted to reading the codebase and writing approved markdown files within a specific directory structure under openspec/changes/.- [PROMPT_INJECTION]: Instructions focus on maintaining interview semantics and user language preferences without attempting to bypass agent safety filters or override system-level instructions.

grill