The Agent Skills Directory

Operational Scope: The skill functions exclusively through markdown instructions and reference text. It does not include any scripts, binary executables, or commands that interact with the host system or external network.
Internal Content References: Supporting material is provided through a local reference file (references/pov-patterns.md). This method for providing additional context is a secure alternative to fetching remote content or using external dependencies.
Indirect Prompt Injection Surface: The skill is designed to ingest and process user-supplied topics, audiences, and tones to generate its output. While this constitutes an ingestion point for untrusted data, the associated risk is negligible due to the lack of privileged capabilities.
Ingestion points: User input for Topic, Audience, and Stance defined in the Workflow section of SKILL.md.
Boundary markers: The skill does not implement specific delimiters or instructions to isolate user-provided data from its own core instructions.
Capability inventory: The skill lacks access to tools, subprocesses, file-writing utilities, or network APIs; its output is strictly limited to natural language text provided to the user.
Sanitization: No explicit data sanitization or filtering is employed for the user-provided inputs.
Context: The risk factor is low because any potential injection within the user input would not have access to sensitive tools or persistent storage through this skill.

pov-line-generator