The Agent Skills Directory

[SAFE]: The skill operates strictly within the local filesystem and the project's documentation directories. It does not perform any network operations, external downloads, or exfiltration of data.
[SAFE]: A manual security checkpoint is enforced via a blocking gate ([STOP — BLOCKING]), which requires the agent to present its draft for user approval before writing any final design artifacts to the disk. This ensures human oversight of the agent's actions.
[SAFE]: The execution environment is further restricted by the disable-model-invocation: true directive, which prevents the agent from autonomously calling tools without explicit instruction.
[NO_CODE]: The skill does not include or execute any scripts, binaries, or third-party dependencies; it consists entirely of instructional markdown and organizational logic for the AI agent.

recipe-blueprint