The Agent Skills Directory

[SAFE]: The skill defines a clear methodology for analyzing visual content using multimodal capabilities. It focuses on objective documentation of visible elements.
[DATA_EXPOSURE]: While the skill uses the Read tool to access image files, this is restricted to the intended task of summarization. No access to sensitive system files or environment variables was detected.
[PROMPT_INJECTION]: There are no instructions designed to bypass agent safety filters or override system-level constraints. The instructional imperatives ('MUST', 'Required behaviors') are legitimate guidelines for the task.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted visual data (screenshots and diagrams) which could contain malicious text. However, it includes robust mitigations via 'Fidelity Rules' that instruct the agent to only describe what is visible and avoid speculating or inferring intent, which prevents the execution of instructions embedded in the images.
[EXTERNAL_DOWNLOADS]: The skill does not perform any network operations or external downloads.

image-summarization