Auto-learns your visual preferences. Adapts to UI, graphics, video, and any creative work.
Security Analysis
medium confidenceThe skill's requested resources and runtime instructions are consistent with a preference-learning design assistant: it asks for no credentials, installs nothing, and only references its bundled docs — but provenance is unknown and the instructions are somewhat vague about where learned preferences are stored.
Name and description (auto-learns visual preferences) align with what the skill actually contains: instruction-only guidance and two reference files. There are no unrelated env vars, binaries, or install steps requested that would be out of scope for a design-preferences assistant.
SKILL.md stays within the domain of observing user choices/feedback and updating a compact preferences file; it references only bundled docs (dimensions.md, criteria.md). It is vague about the exact data sources ('choices, feedback, and reactions') and where/when updates are written, which could lead to broader data collection depending on how the agent implements 'observe and fill'. It does not instruct reading system config, secrets, or contacting external endpoints.
No install spec and no code files — lowest-risk pattern. Nothing is downloaded or written as part of an install step according to the manifest.
The skill requests no environment variables, credentials, or config paths. There are no disproportionate secret requirements relative to its stated purpose.
The instructions imply updating SKILL.md (or storing compact preference entries). That is reasonable for a personalization skill but does require write access to the agent's skill files or storage. always is false (not forced into every run), but autonomous invocation is allowed by default — combined with write capability this could let the skill adapt without frequent explicit consent. Also, the skill lacks author/homepage metadata, increasing provenance risk.
Guidance
This skill appears internally consistent for learning design preferences and asks for nothing sensitive, but take these precautions before installing: 1) Note the package has no source/homepage — provenance is unknown; prefer skills with identifiable authors. 2) Confirm where learned preferences will be stored and whether the agent has permission to modify SKILL.md or other files; restrict write permissions if you want to prevent persistent changes. 3) If you do not want the skill to adapt autonomously, disable autonomous invocation or monitor changes/logs. 4) Avoid sending private credentials or sensitive images to the skill; it is intended to learn from conversational feedback and examples only. If you want higher assurance, ask the publisher for a repo or contact information before use.
Latest Release
v1.0.0
Initial release - Auto-adaptive design preferences
More by @ivangdavila
Published by @ivangdavila on ClawHub
