The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides multiple reference links to documentation hosted at sosumi.ai (e.g., https://sosumi.ai/documentation/naturallanguage). While the name 'Sosumi' is a historical Apple reference, this domain is not an official or trusted source for software documentation, potentially leading users or agents to untrusted content.
[METADATA_POISONING]: The skill and its reference files (e.g., references/translation-patterns.md) mention requirements for non-existent operating system versions, such as 'iOS 26.4' and 'macOS 26.4'. These deceptive versioning claims provide a false context for the software's execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by processing external natural language text. 1. Ingestion points: tokenizer.string and tagger.string in SKILL.md and references/translation-patterns.md. 2. Boundary markers: Absent; instructions do not advise on delimiters or instruction-ignore warnings for processed text. 3. Capability inventory: Performs tokenization, sentiment analysis, translation, and custom Core ML model predictions. 4. Sanitization: Absent; no validation or escaping is applied to input strings before framework processing.

natural-language