The Agent Skills Directory

[SAFE]: The skill implements a robust human-in-the-loop workflow. It requires explicit user approval at multiple 'lock' checkpoints (Style Brief, Inspiration Board, Hypothesis, Anchor Images, Final Style Lock) before proceeding to the next stage or creating files.
[SAFE]: Data management follows security best practices. The instructions explicitly forbid downloading or packaging third-party inspiration images, requiring instead that the agent only preserve source URLs. It also restricts packaging to user-authorized or approved generated evidence.
[SAFE]: Repository and installation actions are scoped to the local environment and the user's managed workflow. The skill specifically states that the creation process does not authorize commits, pushes, pull requests, or publication without separate, explicit user authorization.
[SAFE]: The references/generated-skill-contract.md ensures that any new skill created by this agent will adhere to safety standards, including meaningful exclusions to prevent accidental routing and strict boundaries between content and style.

image-style-creator