The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes content from external URLs via the WebFetch tool. This is a common pattern for research skills, and this specific skill mitigates the risk by mandating human-in-the-loop confirmation before any requirement is generated or stored.
Ingestion points: External URLs fetched via WebFetch and user-provided descriptions of UI/UX reviews or competitor analysis.
Boundary markers: Present. The skill workflow explicitly defines a 'User Confirmation' phase (Step 4) using the AskUserQuestion tool to verify suggestions before they are converted into requirements.
Capability inventory: The skill uses Read and Write for local documentation files (docs/devdocs/), WebFetch for external content, and AskUserQuestion for user interaction.
Sanitization: Absent. The system relies on the user to review the structured 'Insights' list for accuracy and safety before promoting them to the requirements document.
[EXTERNAL_DOWNLOADS]: The skill uses WebFetch to retrieve content from external documentation and reference sites. This is used for its primary function of research and insight collection.

devdocs-insights