The Agent Skills Directory

[SAFE]: The skill operates purely within the context of text generation and local file storage. It does not attempt to access sensitive system files or credentials.- [COMMAND_EXECUTION]: The skill instructs the agent to save generated content to a specific local path (tasks/prd-[feature-name].md). This file system interaction is legitimate and limited to the skill's documented purpose of document creation.- [DATA_EXFILTRATION]: There are no network-based instructions or hidden tools that could transmit data to external servers.- [PROMPT_INJECTION]: The skill contains natural instructional language intended to guide the agent's behavior for its specific task. There are no attempts to bypass model safety filters or override system-level constraints.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied feature descriptions. While this is an ingestion point for untrusted data, the scope of the skill's capabilities (writing a text document) and the predictable output format minimize the risk of malicious instruction execution.

prd