The Agent Skills Directory

[METADATA_POISONING]: The skill metadata and body text contain significant inconsistencies. The author is listed as 'ClawFu' in the metadata section, which contradicts the registered skill author identity. Additionally, while the primary functionality is image batch processing, the 'What Claude Does vs What You Decide' and 'Skill Boundaries' sections describe video production workflows, storyboard frameworks, and creative direction for video. These discrepancies are misleading regarding the skill's actual capabilities and intent.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted files from user-specified paths, presenting an attack surface for indirect prompt injection through file metadata or content.\n
Ingestion points: CLI path arguments in the resize, compress, remove-bg, watermark, and convert commands within scripts/main.py.\n
Boundary markers: None identified; there are no instructions or delimiters used to separate untrusted data from the agent's instructions.\n
Capability inventory: File system writes and directory creation via PIL.Image.save and pathlib.Path.mkdir in scripts/main.py.\n
Sanitization: Limited to file extension filtering in the get_images function, which does not sanitize the contents of the files being processed.

image-batch