The Agent Skills Directory

[SAFE]: The skill is a purely instructional resource consisting of markdown files and prompt templates. It does not include any executable code or binary dependencies, which minimizes the security risk profile.
[SAFE]: No instances of prompt injection, data exfiltration, or obfuscation were detected. The instructions prioritize model reliability and adherence to specified formats.
[SAFE]: The skill implements safety-positive guidelines for generated content, such as Rule 7 in creative-rewriting and Rule 1 in document-qa, which explicitly prohibit hallucination and external knowledge retrieval.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied context and domain knowledge to generate distilled prompts, creating a potential surface for indirect injection if the source data is malicious.
Ingestion points: User context and intent are extracted to build the prompt sections (SKILL.md Activation Step 1).
Boundary markers: The output architecture uses structured XML delimiters (, , , , ) to isolate instruction components.
Capability inventory: The skill provides text analysis and generation capabilities without tool-use or network access.
Sanitization: No explicit sanitization of user-provided content is mentioned, though the structured output template provides inherent organizational safety.

down-skilling