The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate multi-step process for character consistency in image generation. All tools mentioned (generate_image, import_media, list_models, get_timeline, capture_frame) are standard for the platform's intended use-case.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data, such as user-uploaded photos and identity descriptions, which is a standard surface for indirect prompt injection. However, the instructions focus on technical image parameters (lighting, resolution, anatomy) and include specific boundary instructions for the agent to follow, minimizing the risk of accidental instruction obedience from data.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns or unverifiable dependencies were detected. The skill relies on platform-provided models and tools.
[DATA_EXFILTRATION]: No evidence of data exfiltration or sensitive credential exposure was found. The network operations are limited to standard model invocations and media imports within the platform's ecosystem.
[OBFUSCATION]: The skill contains no obfuscated content, encoded strings, or hidden characters.

character-pipeline