The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external URLs in the references/web-resource-index.md file for background rationale and documentation. Most sources are well-known technology and design documentation sites (e.g., Anthropic, Microsoft, PromptingGuide.ai, and NN/g), while one reference points to a third-party community repository on skills.sh. These resources are fetched only for informational purposes and are not executed as code.
[PROMPT_INJECTION]: The skill processes untrusted input in the form of PROMPT_TEXT, creating a surface for potential indirect prompt injection.
Ingestion points: The PROMPT_TEXT input field defined in SKILL.md is the primary source of untrusted data.
Boundary markers: The instructions do not define explicit delimiters or instructions to encapsulate the content of PROMPT_TEXT to prevent it from being interpreted as instructions by the subagents.
Capability inventory: The skill can perform network fetches (restricted to a predefined list of URLs), read local skill files, and generate structured text outputs; it lacks shell execution or broad file system write capabilities.
Sanitization: No explicit sanitization, filtering, or escaping is applied to the input text before it is analyzed by the subagent pipeline.

prompt-structurer