The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits user-generated prompts and uploaded images to the fal.ai API for processing. The instructions in SKILL.md explicitly mandate that the agent obtains user confirmation for egress and paid API usage. It also provides clear warnings against using sensitive or confidential data, mitigating the risk of unauthorized data exposure.
[EXTERNAL_DOWNLOADS]: The skill relies on the fal-client and requests Python packages to communicate with fal.ai. Generated images are downloaded from the service's content delivery network (CDN). These interactions are restricted to the well-known fal.ai service and are central to the skill's purpose.
[COMMAND_EXECUTION]: The workflow involves running scripts/falgen.py, which is a local utility script included with the skill. The script manages API calls, file uploads/downloads, and session cost tracking. It operates within the user's local environment using provided credentials.
[PROMPT_INJECTION]: The skill processes user descriptions to build image generation prompts. The instructions contain detailed prompt-engineering guidelines and specify the use of boundary markers to manage untrusted data ingestion. No evidence of intentional behavior override or safety bypass instructions was found.

image-studio